[PATCH net-next 0/6] Basic XDP Support for DQO RDA Queue Format

[PATCH net-next 0/6] Basic XDP Support for DQO RDA Queue Format

[Harshitha Ramamurthy]

From: Joshua Washington <joshwash@google.com>

This patch series updates the GVE XDP infrastructure and introduces
XDP_PASS and XDP_DROP support for the DQO RDA queue format.

The infrastructure changes of note include an allocation path refactor
for XDP queues, and a unification of RX buffer sizes across queue
formats.

This patch series will be followed by more patch series to introduce
XDP_TX and XDP_REDIRECT support, as well as zero-copy and multi-buffer
support.

Joshua Washington (6):
  gve: remove xdp_xsk_done and xdp_xsk_wakeup statistics
  gve: introduce config-based allocation for XDP
  gve: update GQ RX to use buf_size
  gve: merge packet buffer size fields
  gve: update XDP allocation path support RX buffer posting
  gve: add XDP DROP and PASS support for DQ

 drivers/net/ethernet/google/gve/gve.h         |  72 ++---
 drivers/net/ethernet/google/gve/gve_adminq.c  |   4 +-
 .../ethernet/google/gve/gve_buffer_mgmt_dqo.c |  18 +-
 drivers/net/ethernet/google/gve/gve_ethtool.c |  30 +-
 drivers/net/ethernet/google/gve/gve_main.c    | 288 ++++--------------
 drivers/net/ethernet/google/gve/gve_rx.c      |  30 +-
 drivers/net/ethernet/google/gve/gve_rx_dqo.c  |  81 ++++-
 drivers/net/ethernet/google/gve/gve_tx.c      |  41 +--
 drivers/net/ethernet/google/gve/gve_tx_dqo.c  |  31 +-
 9 files changed, 250 insertions(+), 345 deletions(-)

-- 
2.49.0.rc1.451.g8f38331e32-goog

[PATCH net-next 1/6] gve: remove xdp_xsk_done and xdp_xsk_wakeup statistics

[Harshitha Ramamurthy]

From: Joshua Washington <joshwash@google.com>

These statistics pollute the hotpath and do not have any real-world use
or meaning.

Reviewed-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: Joshua Washington <joshwash@google.com>
Signed-off-by: Harshitha Ramamurthy <hramamurthy@google.com>
---
 drivers/net/ethernet/google/gve/gve.h         | 2 --
 drivers/net/ethernet/google/gve/gve_ethtool.c | 8 +++-----
 drivers/net/ethernet/google/gve/gve_tx.c      | 8 ++------
 3 files changed, 5 insertions(+), 13 deletions(-)

diff --git a/drivers/net/ethernet/google/gve/gve.h b/drivers/net/ethernet/google/gve/gve.h
index 483c43bab3a9..2064e592dfdd 100644
--- a/drivers/net/ethernet/google/gve/gve.h
+++ b/drivers/net/ethernet/google/gve/gve.h
@@ -613,8 +613,6 @@ struct gve_tx_ring {
 	dma_addr_t complq_bus_dqo; /* dma address of the dqo.compl_ring */
 	struct u64_stats_sync statss; /* sync stats for 32bit archs */
 	struct xsk_buff_pool *xsk_pool;
-	u32 xdp_xsk_wakeup;
-	u32 xdp_xsk_done;
 	u64 xdp_xsk_sent;
 	u64 xdp_xmit;
 	u64 xdp_xmit_errors;
diff --git a/drivers/net/ethernet/google/gve/gve_ethtool.c b/drivers/net/ethernet/google/gve/gve_ethtool.c
index a572f1e05934..bc59b5b4235a 100644
--- a/drivers/net/ethernet/google/gve/gve_ethtool.c
+++ b/drivers/net/ethernet/google/gve/gve_ethtool.c
@@ -63,8 +63,8 @@ static const char gve_gstrings_rx_stats[][ETH_GSTRING_LEN] = {
 static const char gve_gstrings_tx_stats[][ETH_GSTRING_LEN] = {
 	"tx_posted_desc[%u]", "tx_completed_desc[%u]", "tx_consumed_desc[%u]", "tx_bytes[%u]",
 	"tx_wake[%u]", "tx_stop[%u]", "tx_event_counter[%u]",
-	"tx_dma_mapping_error[%u]", "tx_xsk_wakeup[%u]",
-	"tx_xsk_done[%u]", "tx_xsk_sent[%u]", "tx_xdp_xmit[%u]", "tx_xdp_xmit_errors[%u]"
+	"tx_dma_mapping_error[%u]",
+	"tx_xsk_sent[%u]", "tx_xdp_xmit[%u]", "tx_xdp_xmit_errors[%u]"
 };
 
 static const char gve_gstrings_adminq_stats[][ETH_GSTRING_LEN] = {
@@ -417,9 +417,7 @@ gve_get_ethtool_stats(struct net_device *netdev,
 					data[i++] = value;
 				}
 			}
-			/* XDP xsk counters */
-			data[i++] = tx->xdp_xsk_wakeup;
-			data[i++] = tx->xdp_xsk_done;
+			/* XDP counters */
 			do {
 				start = u64_stats_fetch_begin(&priv->tx[ring].statss);
 				data[i] = tx->xdp_xsk_sent;
diff --git a/drivers/net/ethernet/google/gve/gve_tx.c b/drivers/net/ethernet/google/gve/gve_tx.c
index 4350ebd9c2bd..c8c067e18059 100644
--- a/drivers/net/ethernet/google/gve/gve_tx.c
+++ b/drivers/net/ethernet/google/gve/gve_tx.c
@@ -959,14 +959,10 @@ static int gve_xsk_tx(struct gve_priv *priv, struct gve_tx_ring *tx,
 
 	spin_lock(&tx->xdp_lock);
 	while (sent < budget) {
-		if (!gve_can_tx(tx, GVE_TX_START_THRESH))
+		if (!gve_can_tx(tx, GVE_TX_START_THRESH) ||
+		    !xsk_tx_peek_desc(tx->xsk_pool, &desc))
 			goto out;
 
-		if (!xsk_tx_peek_desc(tx->xsk_pool, &desc)) {
-			tx->xdp_xsk_done = tx->xdp_xsk_wakeup;
-			goto out;
-		}
-
 		data = xsk_buff_raw_get_data(tx->xsk_pool, desc.addr);
 		nsegs = gve_tx_fill_xdp(priv, tx, data, desc.len, NULL, true);
 		tx->req += nsegs;
-- 
2.49.0.rc1.451.g8f38331e32-goog

[PATCH net-next 2/6] gve: introduce config-based allocation for XDP

[Harshitha Ramamurthy]

From: Joshua Washington <joshwash@google.com>

An earlier patch series[1] introduced RX/TX ring allocation configuration
structs which contained metadata used to allocate and configure new RX
and TX rings. This led to a much cleaner and safer allocation pattern
wherein queue resources were not deallocated until new queue resources
were successfully allocated.

Migrate the XDP allocation path to use the same pattern to allow for the
existence of a single allocation path instead of relying on XDP-specific
allocation methods. These extra allocation methods result in the
duplication of many existing behaviors while being prone to error when
configuration changes unrelated to XDP occur.

Link: https://lore.kernel.org/netdev/20240122182632.1102721-1-shailend@google.com/ [1]
Reviewed-by: Praveen Kaligineedi <pkaligineedi@google.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: Joshua Washington <joshwash@google.com>
Signed-off-by: Harshitha Ramamurthy <hramamurthy@google.com>
---
 drivers/net/ethernet/google/gve/gve.h         |  56 ++--
 drivers/net/ethernet/google/gve/gve_ethtool.c |  19 +-
 drivers/net/ethernet/google/gve/gve_main.c    | 261 ++++--------------
 drivers/net/ethernet/google/gve/gve_rx.c      |   6 +-
 drivers/net/ethernet/google/gve/gve_rx_dqo.c  |   6 +-
 drivers/net/ethernet/google/gve/gve_tx.c      |  33 +--
 drivers/net/ethernet/google/gve/gve_tx_dqo.c  |  31 +--
 7 files changed, 118 insertions(+), 294 deletions(-)

diff --git a/drivers/net/ethernet/google/gve/gve.h b/drivers/net/ethernet/google/gve/gve.h
index 2064e592dfdd..e5cc3fada9c9 100644
--- a/drivers/net/ethernet/google/gve/gve.h
+++ b/drivers/net/ethernet/google/gve/gve.h
@@ -631,10 +631,17 @@ struct gve_notify_block {
 	u32 irq;
 };
 
-/* Tracks allowed and current queue settings */
-struct gve_queue_config {
+/* Tracks allowed and current rx queue settings */
+struct gve_rx_queue_config {
 	u16 max_queues;
-	u16 num_queues; /* current */
+	u16 num_queues;
+};
+
+/* Tracks allowed and current tx queue settings */
+struct gve_tx_queue_config {
+	u16 max_queues;
+	u16 num_queues; /* number of TX queues, excluding XDP queues */
+	u16 num_xdp_queues;
 };
 
 /* Tracks the available and used qpl IDs */
@@ -658,11 +665,11 @@ struct gve_ptype_lut {
 
 /* Parameters for allocating resources for tx queues */
 struct gve_tx_alloc_rings_cfg {
-	struct gve_queue_config *qcfg;
+	struct gve_tx_queue_config *qcfg;
+
+	u16 num_xdp_rings;
 
 	u16 ring_size;
-	u16 start_idx;
-	u16 num_rings;
 	bool raw_addressing;
 
 	/* Allocated resources are returned here */
@@ -672,8 +679,8 @@ struct gve_tx_alloc_rings_cfg {
 /* Parameters for allocating resources for rx queues */
 struct gve_rx_alloc_rings_cfg {
 	/* tx config is also needed to determine QPL ids */
-	struct gve_queue_config *qcfg;
-	struct gve_queue_config *qcfg_tx;
+	struct gve_rx_queue_config *qcfg_rx;
+	struct gve_tx_queue_config *qcfg_tx;
 
 	u16 ring_size;
 	u16 packet_buffer_size;
@@ -764,9 +771,8 @@ struct gve_priv {
 	u32 rx_copybreak; /* copy packets smaller than this */
 	u16 default_num_queues; /* default num queues to set up */
 
-	u16 num_xdp_queues;
-	struct gve_queue_config tx_cfg;
-	struct gve_queue_config rx_cfg;
+	struct gve_tx_queue_config tx_cfg;
+	struct gve_rx_queue_config rx_cfg;
 	u32 num_ntfy_blks; /* spilt between TX and RX so must be even */
 
 	struct gve_registers __iomem *reg_bar0; /* see gve_register.h */
@@ -1039,27 +1045,16 @@ static inline bool gve_is_qpl(struct gve_priv *priv)
 }
 
 /* Returns the number of tx queue page lists */
-static inline u32 gve_num_tx_qpls(const struct gve_queue_config *tx_cfg,
-				  int num_xdp_queues,
+static inline u32 gve_num_tx_qpls(const struct gve_tx_queue_config *tx_cfg,
 				  bool is_qpl)
 {
 	if (!is_qpl)
 		return 0;
-	return tx_cfg->num_queues + num_xdp_queues;
-}
-
-/* Returns the number of XDP tx queue page lists
- */
-static inline u32 gve_num_xdp_qpls(struct gve_priv *priv)
-{
-	if (priv->queue_format != GVE_GQI_QPL_FORMAT)
-		return 0;
-
-	return priv->num_xdp_queues;
+	return tx_cfg->num_queues + tx_cfg->num_xdp_queues;
 }
 
 /* Returns the number of rx queue page lists */
-static inline u32 gve_num_rx_qpls(const struct gve_queue_config *rx_cfg,
+static inline u32 gve_num_rx_qpls(const struct gve_rx_queue_config *rx_cfg,
 				  bool is_qpl)
 {
 	if (!is_qpl)
@@ -1077,7 +1072,8 @@ static inline u32 gve_rx_qpl_id(struct gve_priv *priv, int rx_qid)
 	return priv->tx_cfg.max_queues + rx_qid;
 }
 
-static inline u32 gve_get_rx_qpl_id(const struct gve_queue_config *tx_cfg, int rx_qid)
+static inline u32 gve_get_rx_qpl_id(const struct gve_tx_queue_config *tx_cfg,
+				    int rx_qid)
 {
 	return tx_cfg->max_queues + rx_qid;
 }
@@ -1087,7 +1083,7 @@ static inline u32 gve_tx_start_qpl_id(struct gve_priv *priv)
 	return gve_tx_qpl_id(priv, 0);
 }
 
-static inline u32 gve_rx_start_qpl_id(const struct gve_queue_config *tx_cfg)
+static inline u32 gve_rx_start_qpl_id(const struct gve_tx_queue_config *tx_cfg)
 {
 	return gve_get_rx_qpl_id(tx_cfg, 0);
 }
@@ -1118,7 +1114,7 @@ static inline bool gve_is_gqi(struct gve_priv *priv)
 
 static inline u32 gve_num_tx_queues(struct gve_priv *priv)
 {
-	return priv->tx_cfg.num_queues + priv->num_xdp_queues;
+	return priv->tx_cfg.num_queues + priv->tx_cfg.num_xdp_queues;
 }
 
 static inline u32 gve_xdp_tx_queue_id(struct gve_priv *priv, u32 queue_id)
@@ -1234,8 +1230,8 @@ int gve_adjust_config(struct gve_priv *priv,
 		      struct gve_tx_alloc_rings_cfg *tx_alloc_cfg,
 		      struct gve_rx_alloc_rings_cfg *rx_alloc_cfg);
 int gve_adjust_queues(struct gve_priv *priv,
-		      struct gve_queue_config new_rx_config,
-		      struct gve_queue_config new_tx_config,
+		      struct gve_rx_queue_config new_rx_config,
+		      struct gve_tx_queue_config new_tx_config,
 		      bool reset_rss);
 /* flow steering rule */
 int gve_get_flow_rule_entry(struct gve_priv *priv, struct ethtool_rxnfc *cmd);
diff --git a/drivers/net/ethernet/google/gve/gve_ethtool.c b/drivers/net/ethernet/google/gve/gve_ethtool.c
index bc59b5b4235a..a862031ba5d1 100644
--- a/drivers/net/ethernet/google/gve/gve_ethtool.c
+++ b/drivers/net/ethernet/google/gve/gve_ethtool.c
@@ -475,8 +475,8 @@ static int gve_set_channels(struct net_device *netdev,
 			    struct ethtool_channels *cmd)
 {
 	struct gve_priv *priv = netdev_priv(netdev);
-	struct gve_queue_config new_tx_cfg = priv->tx_cfg;
-	struct gve_queue_config new_rx_cfg = priv->rx_cfg;
+	struct gve_tx_queue_config new_tx_cfg = priv->tx_cfg;
+	struct gve_rx_queue_config new_rx_cfg = priv->rx_cfg;
 	struct ethtool_channels old_settings;
 	int new_tx = cmd->tx_count;
 	int new_rx = cmd->rx_count;
@@ -491,10 +491,17 @@ static int gve_set_channels(struct net_device *netdev,
 	if (!new_rx || !new_tx)
 		return -EINVAL;
 
-	if (priv->num_xdp_queues &&
-	    (new_tx != new_rx || (2 * new_tx > priv->tx_cfg.max_queues))) {
-		dev_err(&priv->pdev->dev, "XDP load failed: The number of configured RX queues should be equal to the number of configured TX queues and the number of configured RX/TX queues should be less than or equal to half the maximum number of RX/TX queues");
-		return -EINVAL;
+	if (priv->xdp_prog) {
+		if (new_tx != new_rx ||
+		    (2 * new_tx > priv->tx_cfg.max_queues)) {
+			dev_err(&priv->pdev->dev, "The number of configured RX queues should be equal to the number of configured TX queues and the number of configured RX/TX queues should be less than or equal to half the maximum number of RX/TX queues when XDP program is installed");
+			return -EINVAL;
+		}
+
+		/* One XDP TX queue per RX queue. */
+		new_tx_cfg.num_xdp_queues = new_rx;
+	} else {
+		new_tx_cfg.num_xdp_queues = 0;
 	}
 
 	if (new_rx != priv->rx_cfg.num_queues &&
diff --git a/drivers/net/ethernet/google/gve/gve_main.c b/drivers/net/ethernet/google/gve/gve_main.c
index 6dcdcaf518f4..354f526a9238 100644
--- a/drivers/net/ethernet/google/gve/gve_main.c
+++ b/drivers/net/ethernet/google/gve/gve_main.c
@@ -795,30 +795,13 @@ static struct gve_queue_page_list *gve_rx_get_qpl(struct gve_priv *priv, int idx
 		return rx->dqo.qpl;
 }
 
-static int gve_register_xdp_qpls(struct gve_priv *priv)
-{
-	int start_id;
-	int err;
-	int i;
-
-	start_id = gve_xdp_tx_start_queue_id(priv);
-	for (i = start_id; i < start_id + gve_num_xdp_qpls(priv); i++) {
-		err = gve_register_qpl(priv, gve_tx_get_qpl(priv, i));
-		/* This failure will trigger a reset - no need to clean up */
-		if (err)
-			return err;
-	}
-	return 0;
-}
-
 static int gve_register_qpls(struct gve_priv *priv)
 {
 	int num_tx_qpls, num_rx_qpls;
 	int err;
 	int i;
 
-	num_tx_qpls = gve_num_tx_qpls(&priv->tx_cfg, gve_num_xdp_qpls(priv),
-				      gve_is_qpl(priv));
+	num_tx_qpls = gve_num_tx_qpls(&priv->tx_cfg, gve_is_qpl(priv));
 	num_rx_qpls = gve_num_rx_qpls(&priv->rx_cfg, gve_is_qpl(priv));
 
 	for (i = 0; i < num_tx_qpls; i++) {
@@ -836,30 +819,13 @@ static int gve_register_qpls(struct gve_priv *priv)
 	return 0;
 }
 
-static int gve_unregister_xdp_qpls(struct gve_priv *priv)
-{
-	int start_id;
-	int err;
-	int i;
-
-	start_id = gve_xdp_tx_start_queue_id(priv);
-	for (i = start_id; i < start_id + gve_num_xdp_qpls(priv); i++) {
-		err = gve_unregister_qpl(priv, gve_tx_get_qpl(priv, i));
-		/* This failure will trigger a reset - no need to clean */
-		if (err)
-			return err;
-	}
-	return 0;
-}
-
 static int gve_unregister_qpls(struct gve_priv *priv)
 {
 	int num_tx_qpls, num_rx_qpls;
 	int err;
 	int i;
 
-	num_tx_qpls = gve_num_tx_qpls(&priv->tx_cfg, gve_num_xdp_qpls(priv),
-				      gve_is_qpl(priv));
+	num_tx_qpls = gve_num_tx_qpls(&priv->tx_cfg, gve_is_qpl(priv));
 	num_rx_qpls = gve_num_rx_qpls(&priv->rx_cfg, gve_is_qpl(priv));
 
 	for (i = 0; i < num_tx_qpls; i++) {
@@ -878,27 +844,6 @@ static int gve_unregister_qpls(struct gve_priv *priv)
 	return 0;
 }
 
-static int gve_create_xdp_rings(struct gve_priv *priv)
-{
-	int err;
-
-	err = gve_adminq_create_tx_queues(priv,
-					  gve_xdp_tx_start_queue_id(priv),
-					  priv->num_xdp_queues);
-	if (err) {
-		netif_err(priv, drv, priv->dev, "failed to create %d XDP tx queues\n",
-			  priv->num_xdp_queues);
-		/* This failure will trigger a reset - no need to clean
-		 * up
-		 */
-		return err;
-	}
-	netif_dbg(priv, drv, priv->dev, "created %d XDP tx queues\n",
-		  priv->num_xdp_queues);
-
-	return 0;
-}
-
 static int gve_create_rings(struct gve_priv *priv)
 {
 	int num_tx_queues = gve_num_tx_queues(priv);
@@ -954,7 +899,7 @@ static void init_xdp_sync_stats(struct gve_priv *priv)
 	int i;
 
 	/* Init stats */
-	for (i = start_id; i < start_id + priv->num_xdp_queues; i++) {
+	for (i = start_id; i < start_id + priv->tx_cfg.num_xdp_queues; i++) {
 		int ntfy_idx = gve_tx_idx_to_ntfy(priv, i);
 
 		u64_stats_init(&priv->tx[i].statss);
@@ -979,24 +924,21 @@ static void gve_init_sync_stats(struct gve_priv *priv)
 static void gve_tx_get_curr_alloc_cfg(struct gve_priv *priv,
 				      struct gve_tx_alloc_rings_cfg *cfg)
 {
-	int num_xdp_queues = priv->xdp_prog ? priv->rx_cfg.num_queues : 0;
-
 	cfg->qcfg = &priv->tx_cfg;
 	cfg->raw_addressing = !gve_is_qpl(priv);
 	cfg->ring_size = priv->tx_desc_cnt;
-	cfg->start_idx = 0;
-	cfg->num_rings = priv->tx_cfg.num_queues + num_xdp_queues;
+	cfg->num_xdp_rings = cfg->qcfg->num_xdp_queues;
 	cfg->tx = priv->tx;
 }
 
-static void gve_tx_stop_rings(struct gve_priv *priv, int start_id, int num_rings)
+static void gve_tx_stop_rings(struct gve_priv *priv, int num_rings)
 {
 	int i;
 
 	if (!priv->tx)
 		return;
 
-	for (i = start_id; i < start_id + num_rings; i++) {
+	for (i = 0; i < num_rings; i++) {
 		if (gve_is_gqi(priv))
 			gve_tx_stop_ring_gqi(priv, i);
 		else
@@ -1004,12 +946,11 @@ static void gve_tx_stop_rings(struct gve_priv *priv, int start_id, int num_rings
 	}
 }
 
-static void gve_tx_start_rings(struct gve_priv *priv, int start_id,
-			       int num_rings)
+static void gve_tx_start_rings(struct gve_priv *priv, int num_rings)
 {
 	int i;
 
-	for (i = start_id; i < start_id + num_rings; i++) {
+	for (i = 0; i < num_rings; i++) {
 		if (gve_is_gqi(priv))
 			gve_tx_start_ring_gqi(priv, i);
 		else
@@ -1017,28 +958,6 @@ static void gve_tx_start_rings(struct gve_priv *priv, int start_id,
 	}
 }
 
-static int gve_alloc_xdp_rings(struct gve_priv *priv)
-{
-	struct gve_tx_alloc_rings_cfg cfg = {0};
-	int err = 0;
-
-	if (!priv->num_xdp_queues)
-		return 0;
-
-	gve_tx_get_curr_alloc_cfg(priv, &cfg);
-	cfg.start_idx = gve_xdp_tx_start_queue_id(priv);
-	cfg.num_rings = priv->num_xdp_queues;
-
-	err = gve_tx_alloc_rings_gqi(priv, &cfg);
-	if (err)
-		return err;
-
-	gve_tx_start_rings(priv, cfg.start_idx, cfg.num_rings);
-	init_xdp_sync_stats(priv);
-
-	return 0;
-}
-
 static int gve_queues_mem_alloc(struct gve_priv *priv,
 				struct gve_tx_alloc_rings_cfg *tx_alloc_cfg,
 				struct gve_rx_alloc_rings_cfg *rx_alloc_cfg)
@@ -1069,26 +988,6 @@ static int gve_queues_mem_alloc(struct gve_priv *priv,
 	return err;
 }
 
-static int gve_destroy_xdp_rings(struct gve_priv *priv)
-{
-	int start_id;
-	int err;
-
-	start_id = gve_xdp_tx_start_queue_id(priv);
-	err = gve_adminq_destroy_tx_queues(priv,
-					   start_id,
-					   priv->num_xdp_queues);
-	if (err) {
-		netif_err(priv, drv, priv->dev,
-			  "failed to destroy XDP queues\n");
-		/* This failure will trigger a reset - no need to clean up */
-		return err;
-	}
-	netif_dbg(priv, drv, priv->dev, "destroyed XDP queues\n");
-
-	return 0;
-}
-
 static int gve_destroy_rings(struct gve_priv *priv)
 {
 	int num_tx_queues = gve_num_tx_queues(priv);
@@ -1113,20 +1012,6 @@ static int gve_destroy_rings(struct gve_priv *priv)
 	return 0;
 }
 
-static void gve_free_xdp_rings(struct gve_priv *priv)
-{
-	struct gve_tx_alloc_rings_cfg cfg = {0};
-
-	gve_tx_get_curr_alloc_cfg(priv, &cfg);
-	cfg.start_idx = gve_xdp_tx_start_queue_id(priv);
-	cfg.num_rings = priv->num_xdp_queues;
-
-	if (priv->tx) {
-		gve_tx_stop_rings(priv, cfg.start_idx, cfg.num_rings);
-		gve_tx_free_rings_gqi(priv, &cfg);
-	}
-}
-
 static void gve_queues_mem_free(struct gve_priv *priv,
 				struct gve_tx_alloc_rings_cfg *tx_cfg,
 				struct gve_rx_alloc_rings_cfg *rx_cfg)
@@ -1253,7 +1138,7 @@ static int gve_reg_xdp_info(struct gve_priv *priv, struct net_device *dev)
 	int i, j;
 	u32 tx_qid;
 
-	if (!priv->num_xdp_queues)
+	if (!priv->tx_cfg.num_xdp_queues)
 		return 0;
 
 	for (i = 0; i < priv->rx_cfg.num_queues; i++) {
@@ -1283,7 +1168,7 @@ static int gve_reg_xdp_info(struct gve_priv *priv, struct net_device *dev)
 		}
 	}
 
-	for (i = 0; i < priv->num_xdp_queues; i++) {
+	for (i = 0; i < priv->tx_cfg.num_xdp_queues; i++) {
 		tx_qid = gve_xdp_tx_queue_id(priv, i);
 		priv->tx[tx_qid].xsk_pool = xsk_get_pool_from_qid(dev, i);
 	}
@@ -1304,7 +1189,7 @@ static void gve_unreg_xdp_info(struct gve_priv *priv)
 {
 	int i, tx_qid;
 
-	if (!priv->num_xdp_queues)
+	if (!priv->tx_cfg.num_xdp_queues || !priv->rx || !priv->tx)
 		return;
 
 	for (i = 0; i < priv->rx_cfg.num_queues; i++) {
@@ -1317,7 +1202,7 @@ static void gve_unreg_xdp_info(struct gve_priv *priv)
 		}
 	}
 
-	for (i = 0; i < priv->num_xdp_queues; i++) {
+	for (i = 0; i < priv->tx_cfg.num_xdp_queues; i++) {
 		tx_qid = gve_xdp_tx_queue_id(priv, i);
 		priv->tx[tx_qid].xsk_pool = NULL;
 	}
@@ -1334,7 +1219,7 @@ static void gve_drain_page_cache(struct gve_priv *priv)
 static void gve_rx_get_curr_alloc_cfg(struct gve_priv *priv,
 				      struct gve_rx_alloc_rings_cfg *cfg)
 {
-	cfg->qcfg = &priv->rx_cfg;
+	cfg->qcfg_rx = &priv->rx_cfg;
 	cfg->qcfg_tx = &priv->tx_cfg;
 	cfg->raw_addressing = !gve_is_qpl(priv);
 	cfg->enable_header_split = priv->header_split_enabled;
@@ -1415,17 +1300,13 @@ static int gve_queues_start(struct gve_priv *priv,
 
 	/* Record new configs into priv */
 	priv->tx_cfg = *tx_alloc_cfg->qcfg;
-	priv->rx_cfg = *rx_alloc_cfg->qcfg;
+	priv->tx_cfg.num_xdp_queues = tx_alloc_cfg->num_xdp_rings;
+	priv->rx_cfg = *rx_alloc_cfg->qcfg_rx;
 	priv->tx_desc_cnt = tx_alloc_cfg->ring_size;
 	priv->rx_desc_cnt = rx_alloc_cfg->ring_size;
 
-	if (priv->xdp_prog)
-		priv->num_xdp_queues = priv->rx_cfg.num_queues;
-	else
-		priv->num_xdp_queues = 0;
-
-	gve_tx_start_rings(priv, 0, tx_alloc_cfg->num_rings);
-	gve_rx_start_rings(priv, rx_alloc_cfg->qcfg->num_queues);
+	gve_tx_start_rings(priv, gve_num_tx_queues(priv));
+	gve_rx_start_rings(priv, rx_alloc_cfg->qcfg_rx->num_queues);
 	gve_init_sync_stats(priv);
 
 	err = netif_set_real_num_tx_queues(dev, priv->tx_cfg.num_queues);
@@ -1477,7 +1358,7 @@ static int gve_queues_start(struct gve_priv *priv,
 	/* return the original error */
 	return err;
 stop_and_free_rings:
-	gve_tx_stop_rings(priv, 0, gve_num_tx_queues(priv));
+	gve_tx_stop_rings(priv, gve_num_tx_queues(priv));
 	gve_rx_stop_rings(priv, priv->rx_cfg.num_queues);
 	gve_queues_mem_remove(priv);
 	return err;
@@ -1526,7 +1407,7 @@ static int gve_queues_stop(struct gve_priv *priv)
 
 	gve_unreg_xdp_info(priv);
 
-	gve_tx_stop_rings(priv, 0, gve_num_tx_queues(priv));
+	gve_tx_stop_rings(priv, gve_num_tx_queues(priv));
 	gve_rx_stop_rings(priv, priv->rx_cfg.num_queues);
 
 	priv->interface_down_cnt++;
@@ -1556,56 +1437,6 @@ static int gve_close(struct net_device *dev)
 	return 0;
 }
 
-static int gve_remove_xdp_queues(struct gve_priv *priv)
-{
-	int err;
-
-	err = gve_destroy_xdp_rings(priv);
-	if (err)
-		return err;
-
-	err = gve_unregister_xdp_qpls(priv);
-	if (err)
-		return err;
-
-	gve_unreg_xdp_info(priv);
-	gve_free_xdp_rings(priv);
-
-	priv->num_xdp_queues = 0;
-	return 0;
-}
-
-static int gve_add_xdp_queues(struct gve_priv *priv)
-{
-	int err;
-
-	priv->num_xdp_queues = priv->rx_cfg.num_queues;
-
-	err = gve_alloc_xdp_rings(priv);
-	if (err)
-		goto err;
-
-	err = gve_reg_xdp_info(priv, priv->dev);
-	if (err)
-		goto free_xdp_rings;
-
-	err = gve_register_xdp_qpls(priv);
-	if (err)
-		goto free_xdp_rings;
-
-	err = gve_create_xdp_rings(priv);
-	if (err)
-		goto free_xdp_rings;
-
-	return 0;
-
-free_xdp_rings:
-	gve_free_xdp_rings(priv);
-err:
-	priv->num_xdp_queues = 0;
-	return err;
-}
-
 static void gve_handle_link_status(struct gve_priv *priv, bool link_status)
 {
 	if (!gve_get_napi_enabled(priv))
@@ -1623,6 +1454,18 @@ static void gve_handle_link_status(struct gve_priv *priv, bool link_status)
 	}
 }
 
+static int gve_configure_rings_xdp(struct gve_priv *priv,
+				   u16 num_xdp_rings)
+{
+	struct gve_tx_alloc_rings_cfg tx_alloc_cfg = {0};
+	struct gve_rx_alloc_rings_cfg rx_alloc_cfg = {0};
+
+	gve_get_curr_alloc_cfgs(priv, &tx_alloc_cfg, &rx_alloc_cfg);
+	tx_alloc_cfg.num_xdp_rings = num_xdp_rings;
+
+	return gve_adjust_config(priv, &tx_alloc_cfg, &rx_alloc_cfg);
+}
+
 static int gve_set_xdp(struct gve_priv *priv, struct bpf_prog *prog,
 		       struct netlink_ext_ack *extack)
 {
@@ -1635,29 +1478,26 @@ static int gve_set_xdp(struct gve_priv *priv, struct bpf_prog *prog,
 		WRITE_ONCE(priv->xdp_prog, prog);
 		if (old_prog)
 			bpf_prog_put(old_prog);
+
+		/* Update priv XDP queue configuration */
+		priv->tx_cfg.num_xdp_queues = priv->xdp_prog ?
+			priv->rx_cfg.num_queues : 0;
 		return 0;
 	}
 
-	gve_turndown(priv);
-	if (!old_prog && prog) {
-		// Allocate XDP TX queues if an XDP program is
-		// being installed
-		err = gve_add_xdp_queues(priv);
-		if (err)
-			goto out;
-	} else if (old_prog && !prog) {
-		// Remove XDP TX queues if an XDP program is
-		// being uninstalled
-		err = gve_remove_xdp_queues(priv);
-		if (err)
-			goto out;
-	}
+	if (!old_prog && prog)
+		err = gve_configure_rings_xdp(priv, priv->rx_cfg.num_queues);
+	else if (old_prog && !prog)
+		err = gve_configure_rings_xdp(priv, 0);
+
+	if (err)
+		goto out;
+
 	WRITE_ONCE(priv->xdp_prog, prog);
 	if (old_prog)
 		bpf_prog_put(old_prog);
 
 out:
-	gve_turnup(priv);
 	status = ioread32be(&priv->reg_bar0->device_status);
 	gve_handle_link_status(priv, GVE_DEVICE_STATUS_LINK_STATUS_MASK & status);
 	return err;
@@ -1908,13 +1748,12 @@ int gve_adjust_config(struct gve_priv *priv,
 }
 
 int gve_adjust_queues(struct gve_priv *priv,
-		      struct gve_queue_config new_rx_config,
-		      struct gve_queue_config new_tx_config,
+		      struct gve_rx_queue_config new_rx_config,
+		      struct gve_tx_queue_config new_tx_config,
 		      bool reset_rss)
 {
 	struct gve_tx_alloc_rings_cfg tx_alloc_cfg = {0};
 	struct gve_rx_alloc_rings_cfg rx_alloc_cfg = {0};
-	int num_xdp_queues;
 	int err;
 
 	gve_get_curr_alloc_cfgs(priv, &tx_alloc_cfg, &rx_alloc_cfg);
@@ -1922,13 +1761,8 @@ int gve_adjust_queues(struct gve_priv *priv,
 	/* Relay the new config from ethtool */
 	tx_alloc_cfg.qcfg = &new_tx_config;
 	rx_alloc_cfg.qcfg_tx = &new_tx_config;
-	rx_alloc_cfg.qcfg = &new_rx_config;
+	rx_alloc_cfg.qcfg_rx = &new_rx_config;
 	rx_alloc_cfg.reset_rss = reset_rss;
-	tx_alloc_cfg.num_rings = new_tx_config.num_queues;
-
-	/* Add dedicated XDP TX queues if enabled. */
-	num_xdp_queues = priv->xdp_prog ? new_rx_config.num_queues : 0;
-	tx_alloc_cfg.num_rings += num_xdp_queues;
 
 	if (netif_running(priv->dev)) {
 		err = gve_adjust_config(priv, &tx_alloc_cfg, &rx_alloc_cfg);
@@ -2056,7 +1890,7 @@ static void gve_turnup(struct gve_priv *priv)
 		napi_schedule(&block->napi);
 	}
 
-	if (priv->num_xdp_queues && gve_supports_xdp_xmit(priv))
+	if (priv->tx_cfg.num_xdp_queues && gve_supports_xdp_xmit(priv))
 		xdp_features_set_redirect_target(priv->dev, false);
 
 	gve_set_napi_enabled(priv);
@@ -2412,6 +2246,7 @@ static int gve_init_priv(struct gve_priv *priv, bool skip_describe_device)
 		priv->rx_cfg.num_queues = min_t(int, priv->default_num_queues,
 						priv->rx_cfg.num_queues);
 	}
+	priv->tx_cfg.num_xdp_queues = 0;
 
 	dev_info(&priv->pdev->dev, "TX queues %d, RX queues %d\n",
 		 priv->tx_cfg.num_queues, priv->rx_cfg.num_queues);
diff --git a/drivers/net/ethernet/google/gve/gve_rx.c b/drivers/net/ethernet/google/gve/gve_rx.c
index acb73d4d0de6..7b774cc510cc 100644
--- a/drivers/net/ethernet/google/gve/gve_rx.c
+++ b/drivers/net/ethernet/google/gve/gve_rx.c
@@ -385,12 +385,12 @@ int gve_rx_alloc_rings_gqi(struct gve_priv *priv,
 	int err = 0;
 	int i, j;
 
-	rx = kvcalloc(cfg->qcfg->max_queues, sizeof(struct gve_rx_ring),
+	rx = kvcalloc(cfg->qcfg_rx->max_queues, sizeof(struct gve_rx_ring),
 		      GFP_KERNEL);
 	if (!rx)
 		return -ENOMEM;
 
-	for (i = 0; i < cfg->qcfg->num_queues; i++) {
+	for (i = 0; i < cfg->qcfg_rx->num_queues; i++) {
 		err = gve_rx_alloc_ring_gqi(priv, cfg, &rx[i], i);
 		if (err) {
 			netif_err(priv, drv, priv->dev,
@@ -419,7 +419,7 @@ void gve_rx_free_rings_gqi(struct gve_priv *priv,
 	if (!rx)
 		return;
 
-	for (i = 0; i < cfg->qcfg->num_queues;  i++)
+	for (i = 0; i < cfg->qcfg_rx->num_queues;  i++)
 		gve_rx_free_ring_gqi(priv, &rx[i], cfg);
 
 	kvfree(rx);
diff --git a/drivers/net/ethernet/google/gve/gve_rx_dqo.c b/drivers/net/ethernet/google/gve/gve_rx_dqo.c
index 856ade0c209f..dcdad6d09bf3 100644
--- a/drivers/net/ethernet/google/gve/gve_rx_dqo.c
+++ b/drivers/net/ethernet/google/gve/gve_rx_dqo.c
@@ -299,12 +299,12 @@ int gve_rx_alloc_rings_dqo(struct gve_priv *priv,
 	int err;
 	int i;
 
-	rx = kvcalloc(cfg->qcfg->max_queues, sizeof(struct gve_rx_ring),
+	rx = kvcalloc(cfg->qcfg_rx->max_queues, sizeof(struct gve_rx_ring),
 		      GFP_KERNEL);
 	if (!rx)
 		return -ENOMEM;
 
-	for (i = 0; i < cfg->qcfg->num_queues; i++) {
+	for (i = 0; i < cfg->qcfg_rx->num_queues; i++) {
 		err = gve_rx_alloc_ring_dqo(priv, cfg, &rx[i], i);
 		if (err) {
 			netif_err(priv, drv, priv->dev,
@@ -333,7 +333,7 @@ void gve_rx_free_rings_dqo(struct gve_priv *priv,
 	if (!rx)
 		return;
 
-	for (i = 0; i < cfg->qcfg->num_queues;  i++)
+	for (i = 0; i < cfg->qcfg_rx->num_queues;  i++)
 		gve_rx_free_ring_dqo(priv, &rx[i], cfg);
 
 	kvfree(rx);
diff --git a/drivers/net/ethernet/google/gve/gve_tx.c b/drivers/net/ethernet/google/gve/gve_tx.c
index c8c067e18059..1b40bf0c811a 100644
--- a/drivers/net/ethernet/google/gve/gve_tx.c
+++ b/drivers/net/ethernet/google/gve/gve_tx.c
@@ -334,27 +334,23 @@ int gve_tx_alloc_rings_gqi(struct gve_priv *priv,
 			   struct gve_tx_alloc_rings_cfg *cfg)
 {
 	struct gve_tx_ring *tx = cfg->tx;
+	int total_queues;
 	int err = 0;
 	int i, j;
 
-	if (cfg->start_idx + cfg->num_rings > cfg->qcfg->max_queues) {
+	total_queues = cfg->qcfg->num_queues + cfg->num_xdp_rings;
+	if (total_queues > cfg->qcfg->max_queues) {
 		netif_err(priv, drv, priv->dev,
 			  "Cannot alloc more than the max num of Tx rings\n");
 		return -EINVAL;
 	}
 
-	if (cfg->start_idx == 0) {
-		tx = kvcalloc(cfg->qcfg->max_queues, sizeof(struct gve_tx_ring),
-			      GFP_KERNEL);
-		if (!tx)
-			return -ENOMEM;
-	} else if (!tx) {
-		netif_err(priv, drv, priv->dev,
-			  "Cannot alloc tx rings from a nonzero start idx without tx array\n");
-		return -EINVAL;
-	}
+	tx = kvcalloc(cfg->qcfg->max_queues, sizeof(struct gve_tx_ring),
+		      GFP_KERNEL);
+	if (!tx)
+		return -ENOMEM;
 
-	for (i = cfg->start_idx; i < cfg->start_idx + cfg->num_rings; i++) {
+	for (i = 0; i < total_queues; i++) {
 		err = gve_tx_alloc_ring_gqi(priv, cfg, &tx[i], i);
 		if (err) {
 			netif_err(priv, drv, priv->dev,
@@ -370,8 +366,7 @@ int gve_tx_alloc_rings_gqi(struct gve_priv *priv,
 cleanup:
 	for (j = 0; j < i; j++)
 		gve_tx_free_ring_gqi(priv, &tx[j], cfg);
-	if (cfg->start_idx == 0)
-		kvfree(tx);
+	kvfree(tx);
 	return err;
 }
 
@@ -384,13 +379,11 @@ void gve_tx_free_rings_gqi(struct gve_priv *priv,
 	if (!tx)
 		return;
 
-	for (i = cfg->start_idx; i < cfg->start_idx + cfg->num_rings; i++)
+	for (i = 0; i < cfg->qcfg->num_queues + cfg->qcfg->num_xdp_queues; i++)
 		gve_tx_free_ring_gqi(priv, &tx[i], cfg);
 
-	if (cfg->start_idx == 0) {
-		kvfree(tx);
-		cfg->tx = NULL;
-	}
+	kvfree(tx);
+	cfg->tx = NULL;
 }
 
 /* gve_tx_avail - Calculates the number of slots available in the ring
@@ -844,7 +837,7 @@ int gve_xdp_xmit(struct net_device *dev, int n, struct xdp_frame **frames,
 		return -ENETDOWN;
 
 	qid = gve_xdp_tx_queue_id(priv,
-				  smp_processor_id() % priv->num_xdp_queues);
+				  smp_processor_id() % priv->tx_cfg.num_xdp_queues);
 
 	tx = &priv->tx[qid];
 
diff --git a/drivers/net/ethernet/google/gve/gve_tx_dqo.c b/drivers/net/ethernet/google/gve/gve_tx_dqo.c
index 394debc62268..2eba868d8037 100644
--- a/drivers/net/ethernet/google/gve/gve_tx_dqo.c
+++ b/drivers/net/ethernet/google/gve/gve_tx_dqo.c
@@ -379,27 +379,23 @@ int gve_tx_alloc_rings_dqo(struct gve_priv *priv,
 			   struct gve_tx_alloc_rings_cfg *cfg)
 {
 	struct gve_tx_ring *tx = cfg->tx;
+	int total_queues;
 	int err = 0;
 	int i, j;
 
-	if (cfg->start_idx + cfg->num_rings > cfg->qcfg->max_queues) {
+	total_queues = cfg->qcfg->num_queues + cfg->num_xdp_rings;
+	if (total_queues > cfg->qcfg->max_queues) {
 		netif_err(priv, drv, priv->dev,
 			  "Cannot alloc more than the max num of Tx rings\n");
 		return -EINVAL;
 	}
 
-	if (cfg->start_idx == 0) {
-		tx = kvcalloc(cfg->qcfg->max_queues, sizeof(struct gve_tx_ring),
-			      GFP_KERNEL);
-		if (!tx)
-			return -ENOMEM;
-	} else if (!tx) {
-		netif_err(priv, drv, priv->dev,
-			  "Cannot alloc tx rings from a nonzero start idx without tx array\n");
-		return -EINVAL;
-	}
+	tx = kvcalloc(cfg->qcfg->max_queues, sizeof(struct gve_tx_ring),
+		      GFP_KERNEL);
+	if (!tx)
+		return -ENOMEM;
 
-	for (i = cfg->start_idx; i < cfg->start_idx + cfg->num_rings; i++) {
+	for (i = 0; i < total_queues; i++) {
 		err = gve_tx_alloc_ring_dqo(priv, cfg, &tx[i], i);
 		if (err) {
 			netif_err(priv, drv, priv->dev,
@@ -415,8 +411,7 @@ int gve_tx_alloc_rings_dqo(struct gve_priv *priv,
 err:
 	for (j = 0; j < i; j++)
 		gve_tx_free_ring_dqo(priv, &tx[j], cfg);
-	if (cfg->start_idx == 0)
-		kvfree(tx);
+	kvfree(tx);
 	return err;
 }
 
@@ -429,13 +424,11 @@ void gve_tx_free_rings_dqo(struct gve_priv *priv,
 	if (!tx)
 		return;
 
-	for (i = cfg->start_idx; i < cfg->start_idx + cfg->num_rings; i++)
+	for (i = 0; i < cfg->qcfg->num_queues + cfg->qcfg->num_xdp_queues; i++)
 		gve_tx_free_ring_dqo(priv, &tx[i], cfg);
 
-	if (cfg->start_idx == 0) {
-		kvfree(tx);
-		cfg->tx = NULL;
-	}
+	kvfree(tx);
+	cfg->tx = NULL;
 }
 
 /* Returns the number of slots available in the ring */
-- 
2.49.0.rc1.451.g8f38331e32-goog

[PATCH net-next 3/6] gve: update GQ RX to use buf_size

[Harshitha Ramamurthy]

From: Joshua Washington <joshwash@google.com>

Commit ebdfae0d377b ("gve: adopt page pool for DQ RDA mode") introduced
a buf_size field to the gve_rx_slot_page_info struct, which can be used
in the datapath to take the place of the packet_buffer_size field, as it
will already be hot in the cache due to its extensive use. Using the
buf_size field in the datapath frees up the packet_buffer_size field in
the GQ-specific RX cacheline to be generalized for GQ and DQ (in the
next patch), as there is currently no common packet buffer size field
between the two queue formats.

Reviewed-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: Joshua Washington <joshwash@google.com>
Signed-off-by: Harshitha Ramamurthy <hramamurthy@google.com>
---
 drivers/net/ethernet/google/gve/gve_rx.c | 24 +++++++++++++++---------
 1 file changed, 15 insertions(+), 9 deletions(-)

diff --git a/drivers/net/ethernet/google/gve/gve_rx.c b/drivers/net/ethernet/google/gve/gve_rx.c
index 7b774cc510cc..9d444e723fcd 100644
--- a/drivers/net/ethernet/google/gve/gve_rx.c
+++ b/drivers/net/ethernet/google/gve/gve_rx.c
@@ -141,12 +141,15 @@ void gve_rx_free_ring_gqi(struct gve_priv *priv, struct gve_rx_ring *rx,
 	netif_dbg(priv, drv, priv->dev, "freed rx ring %d\n", idx);
 }
 
-static void gve_setup_rx_buffer(struct gve_rx_slot_page_info *page_info,
-			     dma_addr_t addr, struct page *page, __be64 *slot_addr)
+static void gve_setup_rx_buffer(struct gve_rx_ring *rx,
+				struct gve_rx_slot_page_info *page_info,
+				dma_addr_t addr, struct page *page,
+				__be64 *slot_addr)
 {
 	page_info->page = page;
 	page_info->page_offset = 0;
 	page_info->page_address = page_address(page);
+	page_info->buf_size = rx->packet_buffer_size;
 	*slot_addr = cpu_to_be64(addr);
 	/* The page already has 1 ref */
 	page_ref_add(page, INT_MAX - 1);
@@ -171,7 +174,7 @@ static int gve_rx_alloc_buffer(struct gve_priv *priv, struct device *dev,
 		return err;
 	}
 
-	gve_setup_rx_buffer(page_info, dma, page, &data_slot->addr);
+	gve_setup_rx_buffer(rx, page_info, dma, page, &data_slot->addr);
 	return 0;
 }
 
@@ -199,7 +202,8 @@ static int gve_rx_prefill_pages(struct gve_rx_ring *rx,
 			struct page *page = rx->data.qpl->pages[i];
 			dma_addr_t addr = i * PAGE_SIZE;
 
-			gve_setup_rx_buffer(&rx->data.page_info[i], addr, page,
+			gve_setup_rx_buffer(rx, &rx->data.page_info[i], addr,
+					    page,
 					    &rx->data.data_ring[i].qpl_offset);
 			continue;
 		}
@@ -222,6 +226,7 @@ static int gve_rx_prefill_pages(struct gve_rx_ring *rx,
 			rx->qpl_copy_pool[j].page = page;
 			rx->qpl_copy_pool[j].page_offset = 0;
 			rx->qpl_copy_pool[j].page_address = page_address(page);
+			rx->qpl_copy_pool[j].buf_size = rx->packet_buffer_size;
 
 			/* The page already has 1 ref. */
 			page_ref_add(page, INT_MAX - 1);
@@ -283,6 +288,7 @@ int gve_rx_alloc_ring_gqi(struct gve_priv *priv,
 
 	rx->gve = priv;
 	rx->q_num = idx;
+	rx->packet_buffer_size = GVE_DEFAULT_RX_BUFFER_SIZE;
 
 	rx->mask = slots - 1;
 	rx->data.raw_addressing = cfg->raw_addressing;
@@ -351,7 +357,6 @@ int gve_rx_alloc_ring_gqi(struct gve_priv *priv,
 	rx->db_threshold = slots / 2;
 	gve_rx_init_ring_state_gqi(rx);
 
-	rx->packet_buffer_size = GVE_DEFAULT_RX_BUFFER_SIZE;
 	gve_rx_ctx_clear(&rx->ctx);
 
 	return 0;
@@ -590,7 +595,7 @@ static struct sk_buff *gve_rx_copy_to_pool(struct gve_rx_ring *rx,
 	copy_page_info->pad = page_info->pad;
 
 	skb = gve_rx_add_frags(napi, copy_page_info,
-			       rx->packet_buffer_size, len, ctx);
+			       copy_page_info->buf_size, len, ctx);
 	if (unlikely(!skb))
 		return NULL;
 
@@ -630,7 +635,8 @@ gve_rx_qpl(struct device *dev, struct net_device *netdev,
 	 * device.
 	 */
 	if (page_info->can_flip) {
-		skb = gve_rx_add_frags(napi, page_info, rx->packet_buffer_size, len, ctx);
+		skb = gve_rx_add_frags(napi, page_info, page_info->buf_size,
+				       len, ctx);
 		/* No point in recycling if we didn't get the skb */
 		if (skb) {
 			/* Make sure that the page isn't freed. */
@@ -680,7 +686,7 @@ static struct sk_buff *gve_rx_skb(struct gve_priv *priv, struct gve_rx_ring *rx,
 			skb = gve_rx_raw_addressing(&priv->pdev->dev, netdev,
 						    page_info, len, napi,
 						    data_slot,
-						    rx->packet_buffer_size, ctx);
+						    page_info->buf_size, ctx);
 		} else {
 			skb = gve_rx_qpl(&priv->pdev->dev, netdev, rx,
 					 page_info, len, napi, data_slot);
@@ -855,7 +861,7 @@ static void gve_rx(struct gve_rx_ring *rx, netdev_features_t feat,
 		void *old_data;
 		int xdp_act;
 
-		xdp_init_buff(&xdp, rx->packet_buffer_size, &rx->xdp_rxq);
+		xdp_init_buff(&xdp, page_info->buf_size, &rx->xdp_rxq);
 		xdp_prepare_buff(&xdp, page_info->page_address +
 				 page_info->page_offset, GVE_RX_PAD,
 				 len, false);
-- 
2.49.0.rc1.451.g8f38331e32-goog

[PATCH net-next 4/6] gve: merge packet buffer size fields

[Harshitha Ramamurthy]

From: Joshua Washington <joshwash@google.com>

The data_buffer_size_dqo field in gve_priv and the packet_buffer_size
field in gve_rx_ring theoretically have the same meaning, but they are
defined in two different places and used in two separate contexts. There
is no good reason for this, so this change merges those fields into the
packet_buffer_size field in the RX ring.

This change also introduces a packet_buffer_size field to struct
gve_rx_queue_config to account for cases where queues are not allocated,
such as when the interface is down.

Reviewed-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: Joshua Washington <joshwash@google.com>
Signed-off-by: Harshitha Ramamurthy <hramamurthy@google.com>
---
 drivers/net/ethernet/google/gve/gve.h                 | 4 ++--
 drivers/net/ethernet/google/gve/gve_adminq.c          | 4 +---
 drivers/net/ethernet/google/gve/gve_buffer_mgmt_dqo.c | 7 +++----
 drivers/net/ethernet/google/gve/gve_ethtool.c         | 3 +--
 drivers/net/ethernet/google/gve/gve_main.c            | 8 +++-----
 drivers/net/ethernet/google/gve/gve_rx.c              | 2 +-
 drivers/net/ethernet/google/gve/gve_rx_dqo.c          | 1 +
 7 files changed, 12 insertions(+), 17 deletions(-)

diff --git a/drivers/net/ethernet/google/gve/gve.h b/drivers/net/ethernet/google/gve/gve.h
index e5cc3fada9c9..9895541eddae 100644
--- a/drivers/net/ethernet/google/gve/gve.h
+++ b/drivers/net/ethernet/google/gve/gve.h
@@ -227,6 +227,7 @@ struct gve_rx_cnts {
 /* Contains datapath state used to represent an RX queue. */
 struct gve_rx_ring {
 	struct gve_priv *gve;
+	u16 packet_buffer_size;
 	union {
 		/* GQI fields */
 		struct {
@@ -235,7 +236,6 @@ struct gve_rx_ring {
 
 			/* threshold for posting new buffs and descs */
 			u32 db_threshold;
-			u16 packet_buffer_size;
 
 			u32 qpl_copy_pool_mask;
 			u32 qpl_copy_pool_head;
@@ -635,6 +635,7 @@ struct gve_notify_block {
 struct gve_rx_queue_config {
 	u16 max_queues;
 	u16 num_queues;
+	u16 packet_buffer_size;
 };
 
 /* Tracks allowed and current tx queue settings */
@@ -842,7 +843,6 @@ struct gve_priv {
 	struct gve_ptype_lut *ptype_lut_dqo;
 
 	/* Must be a power of two. */
-	u16 data_buffer_size_dqo;
 	u16 max_rx_buffer_size; /* device limit */
 
 	enum gve_queue_format queue_format;
diff --git a/drivers/net/ethernet/google/gve/gve_adminq.c b/drivers/net/ethernet/google/gve/gve_adminq.c
index be7a423e5ab9..3e8fc33cc11f 100644
--- a/drivers/net/ethernet/google/gve/gve_adminq.c
+++ b/drivers/net/ethernet/google/gve/gve_adminq.c
@@ -731,6 +731,7 @@ static void gve_adminq_get_create_rx_queue_cmd(struct gve_priv *priv,
 		.ntfy_id = cpu_to_be32(rx->ntfy_id),
 		.queue_resources_addr = cpu_to_be64(rx->q_resources_bus),
 		.rx_ring_size = cpu_to_be16(priv->rx_desc_cnt),
+		.packet_buffer_size = cpu_to_be16(rx->packet_buffer_size),
 	};
 
 	if (gve_is_gqi(priv)) {
@@ -743,7 +744,6 @@ static void gve_adminq_get_create_rx_queue_cmd(struct gve_priv *priv,
 			cpu_to_be64(rx->data.data_bus);
 		cmd->create_rx_queue.index = cpu_to_be32(queue_index);
 		cmd->create_rx_queue.queue_page_list_id = cpu_to_be32(qpl_id);
-		cmd->create_rx_queue.packet_buffer_size = cpu_to_be16(rx->packet_buffer_size);
 	} else {
 		u32 qpl_id = 0;
 
@@ -756,8 +756,6 @@ static void gve_adminq_get_create_rx_queue_cmd(struct gve_priv *priv,
 			cpu_to_be64(rx->dqo.complq.bus);
 		cmd->create_rx_queue.rx_data_ring_addr =
 			cpu_to_be64(rx->dqo.bufq.bus);
-		cmd->create_rx_queue.packet_buffer_size =
-			cpu_to_be16(priv->data_buffer_size_dqo);
 		cmd->create_rx_queue.rx_buff_ring_size =
 			cpu_to_be16(priv->rx_desc_cnt);
 		cmd->create_rx_queue.enable_rsc =
diff --git a/drivers/net/ethernet/google/gve/gve_buffer_mgmt_dqo.c b/drivers/net/ethernet/google/gve/gve_buffer_mgmt_dqo.c
index af84cb88f828..f9824664d04c 100644
--- a/drivers/net/ethernet/google/gve/gve_buffer_mgmt_dqo.c
+++ b/drivers/net/ethernet/google/gve/gve_buffer_mgmt_dqo.c
@@ -139,7 +139,7 @@ int gve_alloc_qpl_page_dqo(struct gve_rx_ring *rx,
 	buf_state->page_info.page_offset = 0;
 	buf_state->page_info.page_address =
 		page_address(buf_state->page_info.page);
-	buf_state->page_info.buf_size = priv->data_buffer_size_dqo;
+	buf_state->page_info.buf_size = rx->packet_buffer_size;
 	buf_state->last_single_ref_offset = 0;
 
 	/* The page already has 1 ref. */
@@ -162,7 +162,7 @@ void gve_free_qpl_page_dqo(struct gve_rx_buf_state_dqo *buf_state)
 void gve_try_recycle_buf(struct gve_priv *priv, struct gve_rx_ring *rx,
 			 struct gve_rx_buf_state_dqo *buf_state)
 {
-	const u16 data_buffer_size = priv->data_buffer_size_dqo;
+	const u16 data_buffer_size = rx->packet_buffer_size;
 	int pagecount;
 
 	/* Can't reuse if we only fit one buffer per page */
@@ -217,10 +217,9 @@ void gve_free_to_page_pool(struct gve_rx_ring *rx,
 static int gve_alloc_from_page_pool(struct gve_rx_ring *rx,
 				    struct gve_rx_buf_state_dqo *buf_state)
 {
-	struct gve_priv *priv = rx->gve;
 	netmem_ref netmem;
 
-	buf_state->page_info.buf_size = priv->data_buffer_size_dqo;
+	buf_state->page_info.buf_size = rx->packet_buffer_size;
 	netmem = page_pool_alloc_netmem(rx->dqo.page_pool,
 					&buf_state->page_info.page_offset,
 					&buf_state->page_info.buf_size,
diff --git a/drivers/net/ethernet/google/gve/gve_ethtool.c b/drivers/net/ethernet/google/gve/gve_ethtool.c
index a862031ba5d1..31a21ccf4863 100644
--- a/drivers/net/ethernet/google/gve/gve_ethtool.c
+++ b/drivers/net/ethernet/google/gve/gve_ethtool.c
@@ -647,8 +647,7 @@ static int gve_set_tunable(struct net_device *netdev,
 	switch (etuna->id) {
 	case ETHTOOL_RX_COPYBREAK:
 	{
-		u32 max_copybreak = gve_is_gqi(priv) ?
-			GVE_DEFAULT_RX_BUFFER_SIZE : priv->data_buffer_size_dqo;
+		u32 max_copybreak = priv->rx_cfg.packet_buffer_size;
 
 		len = *(u32 *)value;
 		if (len > max_copybreak)
diff --git a/drivers/net/ethernet/google/gve/gve_main.c b/drivers/net/ethernet/google/gve/gve_main.c
index 354f526a9238..20aabbe0e518 100644
--- a/drivers/net/ethernet/google/gve/gve_main.c
+++ b/drivers/net/ethernet/google/gve/gve_main.c
@@ -1224,9 +1224,7 @@ static void gve_rx_get_curr_alloc_cfg(struct gve_priv *priv,
 	cfg->raw_addressing = !gve_is_qpl(priv);
 	cfg->enable_header_split = priv->header_split_enabled;
 	cfg->ring_size = priv->rx_desc_cnt;
-	cfg->packet_buffer_size = gve_is_gqi(priv) ?
-				  GVE_DEFAULT_RX_BUFFER_SIZE :
-				  priv->data_buffer_size_dqo;
+	cfg->packet_buffer_size = priv->rx_cfg.packet_buffer_size;
 	cfg->rx = priv->rx;
 }
 
@@ -1331,7 +1329,7 @@ static int gve_queues_start(struct gve_priv *priv,
 		goto reset;
 
 	priv->header_split_enabled = rx_alloc_cfg->enable_header_split;
-	priv->data_buffer_size_dqo = rx_alloc_cfg->packet_buffer_size;
+	priv->rx_cfg.packet_buffer_size = rx_alloc_cfg->packet_buffer_size;
 
 	err = gve_create_rings(priv);
 	if (err)
@@ -2627,7 +2625,7 @@ static int gve_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
 	priv->service_task_flags = 0x0;
 	priv->state_flags = 0x0;
 	priv->ethtool_flags = 0x0;
-	priv->data_buffer_size_dqo = GVE_DEFAULT_RX_BUFFER_SIZE;
+	priv->rx_cfg.packet_buffer_size = GVE_DEFAULT_RX_BUFFER_SIZE;
 	priv->max_rx_buffer_size = GVE_DEFAULT_RX_BUFFER_SIZE;
 
 	gve_set_probe_in_progress(priv);
diff --git a/drivers/net/ethernet/google/gve/gve_rx.c b/drivers/net/ethernet/google/gve/gve_rx.c
index 9d444e723fcd..90e875c1832f 100644
--- a/drivers/net/ethernet/google/gve/gve_rx.c
+++ b/drivers/net/ethernet/google/gve/gve_rx.c
@@ -288,7 +288,7 @@ int gve_rx_alloc_ring_gqi(struct gve_priv *priv,
 
 	rx->gve = priv;
 	rx->q_num = idx;
-	rx->packet_buffer_size = GVE_DEFAULT_RX_BUFFER_SIZE;
+	rx->packet_buffer_size = cfg->packet_buffer_size;
 
 	rx->mask = slots - 1;
 	rx->data.raw_addressing = cfg->raw_addressing;
diff --git a/drivers/net/ethernet/google/gve/gve_rx_dqo.c b/drivers/net/ethernet/google/gve/gve_rx_dqo.c
index dcdad6d09bf3..5fbcf93a54e0 100644
--- a/drivers/net/ethernet/google/gve/gve_rx_dqo.c
+++ b/drivers/net/ethernet/google/gve/gve_rx_dqo.c
@@ -223,6 +223,7 @@ int gve_rx_alloc_ring_dqo(struct gve_priv *priv,
 	memset(rx, 0, sizeof(*rx));
 	rx->gve = priv;
 	rx->q_num = idx;
+	rx->packet_buffer_size = cfg->packet_buffer_size;
 
 	rx->dqo.num_buf_states = cfg->raw_addressing ? buffer_queue_slots :
 		gve_get_rx_pages_per_qpl_dqo(cfg->ring_size);
-- 
2.49.0.rc1.451.g8f38331e32-goog

[PATCH net-next 5/6] gve: update XDP allocation path support RX buffer posting

[Harshitha Ramamurthy]

From: Joshua Washington <joshwash@google.com>

In order to support installing an XDP program on DQ, RX buffers need to
be reposted using 4K buffers, which is larger than the default packet
buffer size of 2K. This is needed to accommodate the extra head and tail
that accompanies the data portion of an XDP buffer. Continuing to use 2K
buffers would mean that the packet buffer size for the NIC would have to
be restricted to 2048 - 320 - 256 = 1472B. However, this is problematic
for two reasons: first, 1472 is not a packet buffer size accepted by
GVE; second, at least 1474B of buffer space is needed to accommodate an
MTU of 1460, which is the default on GCP. As such, we allocate 4K
buffers, and post a 2K section of those 4K buffers (offset relative to
the XDP headroom) to the NIC for DMA to avoid a potential extra copy.
Because the GQ-QPL datapath requires copies regardless, this change was
not needed to support XDP in that case.

To capture this subtlety, a new field, packet_buffer_truesize, has been
added to the rx ring struct to represent size of the allocated buffer,
while packet_buffer_size has been left to represent the portion of the
buffer posted to the NIC.

Reviewed-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: Praveen Kaligineedi <pkaligineedi@google.com>
Signed-off-by: Joshua Washington <joshwash@google.com>
Signed-off-by: Harshitha Ramamurthy <hramamurthy@google.com>
---
 drivers/net/ethernet/google/gve/gve.h         | 12 ++++++++--
 .../ethernet/google/gve/gve_buffer_mgmt_dqo.c | 17 +++++++++-----
 drivers/net/ethernet/google/gve/gve_main.c    | 19 +++++++++++++---
 drivers/net/ethernet/google/gve/gve_rx_dqo.c  | 22 ++++++++++++++-----
 4 files changed, 53 insertions(+), 17 deletions(-)

diff --git a/drivers/net/ethernet/google/gve/gve.h b/drivers/net/ethernet/google/gve/gve.h
index 9895541eddae..2fab38c8ee78 100644
--- a/drivers/net/ethernet/google/gve/gve.h
+++ b/drivers/net/ethernet/google/gve/gve.h
@@ -59,6 +59,8 @@
 
 #define GVE_MAX_RX_BUFFER_SIZE 4096
 
+#define GVE_XDP_RX_BUFFER_SIZE_DQO 4096
+
 #define GVE_DEFAULT_RX_BUFFER_OFFSET 2048
 
 #define GVE_PAGE_POOL_SIZE_MULTIPLIER 4
@@ -227,7 +229,11 @@ struct gve_rx_cnts {
 /* Contains datapath state used to represent an RX queue. */
 struct gve_rx_ring {
 	struct gve_priv *gve;
-	u16 packet_buffer_size;
+
+	u16 packet_buffer_size;		/* Size of buffer posted to NIC */
+	u16 packet_buffer_truesize;	/* Total size of RX buffer */
+	u16 rx_headroom;
+
 	union {
 		/* GQI fields */
 		struct {
@@ -688,6 +694,7 @@ struct gve_rx_alloc_rings_cfg {
 	bool raw_addressing;
 	bool enable_header_split;
 	bool reset_rss;
+	bool xdp;
 
 	/* Allocated resources are returned here */
 	struct gve_rx_ring *rx;
@@ -1218,7 +1225,8 @@ void gve_free_buffer(struct gve_rx_ring *rx,
 		     struct gve_rx_buf_state_dqo *buf_state);
 int gve_alloc_buffer(struct gve_rx_ring *rx, struct gve_rx_desc_dqo *desc);
 struct page_pool *gve_rx_create_page_pool(struct gve_priv *priv,
-					  struct gve_rx_ring *rx);
+					  struct gve_rx_ring *rx,
+					  bool xdp);
 
 /* Reset */
 void gve_schedule_reset(struct gve_priv *priv);
diff --git a/drivers/net/ethernet/google/gve/gve_buffer_mgmt_dqo.c b/drivers/net/ethernet/google/gve/gve_buffer_mgmt_dqo.c
index f9824664d04c..a71883e1d920 100644
--- a/drivers/net/ethernet/google/gve/gve_buffer_mgmt_dqo.c
+++ b/drivers/net/ethernet/google/gve/gve_buffer_mgmt_dqo.c
@@ -139,7 +139,8 @@ int gve_alloc_qpl_page_dqo(struct gve_rx_ring *rx,
 	buf_state->page_info.page_offset = 0;
 	buf_state->page_info.page_address =
 		page_address(buf_state->page_info.page);
-	buf_state->page_info.buf_size = rx->packet_buffer_size;
+	buf_state->page_info.buf_size = rx->packet_buffer_truesize;
+	buf_state->page_info.pad = rx->rx_headroom;
 	buf_state->last_single_ref_offset = 0;
 
 	/* The page already has 1 ref. */
@@ -162,7 +163,7 @@ void gve_free_qpl_page_dqo(struct gve_rx_buf_state_dqo *buf_state)
 void gve_try_recycle_buf(struct gve_priv *priv, struct gve_rx_ring *rx,
 			 struct gve_rx_buf_state_dqo *buf_state)
 {
-	const u16 data_buffer_size = rx->packet_buffer_size;
+	const u16 data_buffer_size = rx->packet_buffer_truesize;
 	int pagecount;
 
 	/* Can't reuse if we only fit one buffer per page */
@@ -219,7 +220,7 @@ static int gve_alloc_from_page_pool(struct gve_rx_ring *rx,
 {
 	netmem_ref netmem;
 
-	buf_state->page_info.buf_size = rx->packet_buffer_size;
+	buf_state->page_info.buf_size = rx->packet_buffer_truesize;
 	netmem = page_pool_alloc_netmem(rx->dqo.page_pool,
 					&buf_state->page_info.page_offset,
 					&buf_state->page_info.buf_size,
@@ -231,12 +232,14 @@ static int gve_alloc_from_page_pool(struct gve_rx_ring *rx,
 	buf_state->page_info.netmem = netmem;
 	buf_state->page_info.page_address = netmem_address(netmem);
 	buf_state->addr = page_pool_get_dma_addr_netmem(netmem);
+	buf_state->page_info.pad = rx->dqo.page_pool->p.offset;
 
 	return 0;
 }
 
 struct page_pool *gve_rx_create_page_pool(struct gve_priv *priv,
-					  struct gve_rx_ring *rx)
+					  struct gve_rx_ring *rx,
+					  bool xdp)
 {
 	u32 ntfy_id = gve_rx_idx_to_ntfy(priv, rx->q_num);
 	struct page_pool_params pp = {
@@ -247,7 +250,8 @@ struct page_pool *gve_rx_create_page_pool(struct gve_priv *priv,
 		.netdev = priv->dev,
 		.napi = &priv->ntfy_blocks[ntfy_id].napi,
 		.max_len = PAGE_SIZE,
-		.dma_dir = DMA_FROM_DEVICE,
+		.dma_dir = xdp ? DMA_BIDIRECTIONAL : DMA_FROM_DEVICE,
+		.offset = xdp ? XDP_PACKET_HEADROOM : 0,
 	};
 
 	return page_pool_create(&pp);
@@ -301,7 +305,8 @@ int gve_alloc_buffer(struct gve_rx_ring *rx, struct gve_rx_desc_dqo *desc)
 	}
 	desc->buf_id = cpu_to_le16(buf_state - rx->dqo.buf_states);
 	desc->buf_addr = cpu_to_le64(buf_state->addr +
-				     buf_state->page_info.page_offset);
+				     buf_state->page_info.page_offset +
+				     buf_state->page_info.pad);
 
 	return 0;
 
diff --git a/drivers/net/ethernet/google/gve/gve_main.c b/drivers/net/ethernet/google/gve/gve_main.c
index 20aabbe0e518..cb2f9978f45e 100644
--- a/drivers/net/ethernet/google/gve/gve_main.c
+++ b/drivers/net/ethernet/google/gve/gve_main.c
@@ -1149,8 +1149,14 @@ static int gve_reg_xdp_info(struct gve_priv *priv, struct net_device *dev)
 				       napi->napi_id);
 		if (err)
 			goto err;
-		err = xdp_rxq_info_reg_mem_model(&rx->xdp_rxq,
-						 MEM_TYPE_PAGE_SHARED, NULL);
+		if (gve_is_qpl(priv))
+			err = xdp_rxq_info_reg_mem_model(&rx->xdp_rxq,
+							 MEM_TYPE_PAGE_SHARED,
+							 NULL);
+		else
+			err = xdp_rxq_info_reg_mem_model(&rx->xdp_rxq,
+							 MEM_TYPE_PAGE_POOL,
+							 rx->dqo.page_pool);
 		if (err)
 			goto err;
 		rx->xsk_pool = xsk_get_pool_from_qid(dev, i);
@@ -1226,6 +1232,7 @@ static void gve_rx_get_curr_alloc_cfg(struct gve_priv *priv,
 	cfg->ring_size = priv->rx_desc_cnt;
 	cfg->packet_buffer_size = priv->rx_cfg.packet_buffer_size;
 	cfg->rx = priv->rx;
+	cfg->xdp = !!cfg->qcfg_tx->num_xdp_queues;
 }
 
 void gve_get_curr_alloc_cfgs(struct gve_priv *priv,
@@ -1461,6 +1468,7 @@ static int gve_configure_rings_xdp(struct gve_priv *priv,
 	gve_get_curr_alloc_cfgs(priv, &tx_alloc_cfg, &rx_alloc_cfg);
 	tx_alloc_cfg.num_xdp_rings = num_xdp_rings;
 
+	rx_alloc_cfg.xdp = !!num_xdp_rings;
 	return gve_adjust_config(priv, &tx_alloc_cfg, &rx_alloc_cfg);
 }
 
@@ -1629,6 +1637,7 @@ static int gve_xsk_wakeup(struct net_device *dev, u32 queue_id, u32 flags)
 static int verify_xdp_configuration(struct net_device *dev)
 {
 	struct gve_priv *priv = netdev_priv(dev);
+	u16 max_xdp_mtu;
 
 	if (dev->features & NETIF_F_LRO) {
 		netdev_warn(dev, "XDP is not supported when LRO is on.\n");
@@ -1641,7 +1650,11 @@ static int verify_xdp_configuration(struct net_device *dev)
 		return -EOPNOTSUPP;
 	}
 
-	if (dev->mtu > GVE_DEFAULT_RX_BUFFER_SIZE - sizeof(struct ethhdr) - GVE_RX_PAD) {
+	max_xdp_mtu = priv->rx_cfg.packet_buffer_size - sizeof(struct ethhdr);
+	if (priv->queue_format == GVE_GQI_QPL_FORMAT)
+		max_xdp_mtu -= GVE_RX_PAD;
+
+	if (dev->mtu > max_xdp_mtu) {
 		netdev_warn(dev, "XDP is not supported for mtu %d.\n",
 			    dev->mtu);
 		return -EOPNOTSUPP;
diff --git a/drivers/net/ethernet/google/gve/gve_rx_dqo.c b/drivers/net/ethernet/google/gve/gve_rx_dqo.c
index 5fbcf93a54e0..2edf3c632cbd 100644
--- a/drivers/net/ethernet/google/gve/gve_rx_dqo.c
+++ b/drivers/net/ethernet/google/gve/gve_rx_dqo.c
@@ -225,6 +225,14 @@ int gve_rx_alloc_ring_dqo(struct gve_priv *priv,
 	rx->q_num = idx;
 	rx->packet_buffer_size = cfg->packet_buffer_size;
 
+	if (cfg->xdp) {
+		rx->packet_buffer_truesize = GVE_XDP_RX_BUFFER_SIZE_DQO;
+		rx->rx_headroom = XDP_PACKET_HEADROOM;
+	} else {
+		rx->packet_buffer_truesize = rx->packet_buffer_size;
+		rx->rx_headroom = 0;
+	}
+
 	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(rx->dqo.num_buf_states,
@@ -254,7 +262,7 @@ int gve_rx_alloc_ring_dqo(struct gve_priv *priv,
 		goto err;
 
 	if (cfg->raw_addressing) {
-		pool = gve_rx_create_page_pool(priv, rx);
+		pool = gve_rx_create_page_pool(priv, rx, cfg->xdp);
 		if (IS_ERR(pool))
 			goto err;
 
@@ -484,14 +492,15 @@ static void gve_skb_add_rx_frag(struct gve_rx_ring *rx,
 	if (rx->dqo.page_pool) {
 		skb_add_rx_frag_netmem(rx->ctx.skb_tail, num_frags,
 				       buf_state->page_info.netmem,
-				       buf_state->page_info.page_offset,
-				       buf_len,
+				       buf_state->page_info.page_offset +
+				       buf_state->page_info.pad, buf_len,
 				       buf_state->page_info.buf_size);
 	} else {
 		skb_add_rx_frag(rx->ctx.skb_tail, num_frags,
 				buf_state->page_info.page,
-				buf_state->page_info.page_offset,
-				buf_len, buf_state->page_info.buf_size);
+				buf_state->page_info.page_offset +
+				buf_state->page_info.pad, buf_len,
+				buf_state->page_info.buf_size);
 	}
 }
 
@@ -611,7 +620,8 @@ static int gve_rx_dqo(struct napi_struct *napi, struct gve_rx_ring *rx,
 
 	/* Sync the portion of dma buffer for CPU to read. */
 	dma_sync_single_range_for_cpu(&priv->pdev->dev, buf_state->addr,
-				      buf_state->page_info.page_offset,
+				      buf_state->page_info.page_offset +
+				      buf_state->page_info.pad,
 				      buf_len, DMA_FROM_DEVICE);
 
 	/* Append to current skb if one exists. */
-- 
2.49.0.rc1.451.g8f38331e32-goog

[PATCH net-next 6/6] gve: add XDP DROP and PASS support for DQ

[Harshitha Ramamurthy]

From: Joshua Washington <joshwash@google.com>

This patch adds support for running XDP programs on DQ, along with
rudimentary processing for XDP_DROP and XDP_PASS. These actions require
very limited driver functionality when it comes to processing an XDP
buffer, so currently if the XDP action is not XDP_PASS, the packet is
dropped and stats are updated.

Reviewed-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: Praveen Kaliginedi <pkaligineedi@google.com>
Signed-off-by: Joshua Washington <joshwash@google.com>
Signed-off-by: Harshitha Ramamurthy<hramamurthy@google.com>
---
 drivers/net/ethernet/google/gve/gve_rx_dqo.c | 52 ++++++++++++++++++++
 1 file changed, 52 insertions(+)

diff --git a/drivers/net/ethernet/google/gve/gve_rx_dqo.c b/drivers/net/ethernet/google/gve/gve_rx_dqo.c
index 2edf3c632cbd..deb869eb38e0 100644
--- a/drivers/net/ethernet/google/gve/gve_rx_dqo.c
+++ b/drivers/net/ethernet/google/gve/gve_rx_dqo.c
@@ -546,6 +546,29 @@ static int gve_rx_append_frags(struct napi_struct *napi,
 	return 0;
 }
 
+static void gve_xdp_done_dqo(struct gve_priv *priv, struct gve_rx_ring *rx,
+			     struct xdp_buff *xdp, struct bpf_prog *xprog,
+			     int xdp_act,
+			     struct gve_rx_buf_state_dqo *buf_state)
+{
+	u64_stats_update_begin(&rx->statss);
+	switch (xdp_act) {
+	case XDP_ABORTED:
+	case XDP_DROP:
+	default:
+		rx->xdp_actions[xdp_act]++;
+		break;
+	case XDP_TX:
+		rx->xdp_tx_errors++;
+		break;
+	case XDP_REDIRECT:
+		rx->xdp_redirect_errors++;
+		break;
+	}
+	u64_stats_update_end(&rx->statss);
+	gve_free_buffer(rx, buf_state);
+}
+
 /* Returns 0 if descriptor is completed successfully.
  * Returns -EINVAL if descriptor is invalid.
  * Returns -ENOMEM if data cannot be copied to skb.
@@ -560,6 +583,7 @@ static int gve_rx_dqo(struct napi_struct *napi, struct gve_rx_ring *rx,
 	const bool hsplit = compl_desc->split_header;
 	struct gve_rx_buf_state_dqo *buf_state;
 	struct gve_priv *priv = rx->gve;
+	struct bpf_prog *xprog;
 	u16 buf_len;
 	u16 hdr_len;
 
@@ -633,6 +657,34 @@ static int gve_rx_dqo(struct napi_struct *napi, struct gve_rx_ring *rx,
 		return 0;
 	}
 
+	xprog = READ_ONCE(priv->xdp_prog);
+	if (xprog) {
+		struct xdp_buff xdp;
+		void *old_data;
+		int xdp_act;
+
+		xdp_init_buff(&xdp, buf_state->page_info.buf_size,
+			      &rx->xdp_rxq);
+		xdp_prepare_buff(&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;
+		if (xdp_act != XDP_PASS) {
+			gve_xdp_done_dqo(priv, rx, &xdp, xprog, xdp_act,
+					 buf_state);
+			return 0;
+		}
+
+		u64_stats_update_begin(&rx->statss);
+		rx->xdp_actions[XDP_PASS]++;
+		u64_stats_update_end(&rx->statss);
+	}
+
 	if (eop && buf_len <= priv->rx_copybreak) {
 		rx->ctx.skb_head = gve_rx_copy(priv->dev, napi,
 					       &buf_state->page_info, buf_len);
-- 
2.49.0.rc1.451.g8f38331e32-goog

Re: [PATCH net-next 0/6] Basic XDP Support for DQO RDA Queue Format

[patchwork-bot+netdevbpf]

Hello:

This series was applied to netdev/net-next.git (main)
by Jakub Kicinski <kuba@kernel.org>:

On Fri, 21 Mar 2025 00:29:04 +0000 you wrote:
> From: Joshua Washington <joshwash@google.com>
> 
> This patch series updates the GVE XDP infrastructure and introduces
> XDP_PASS and XDP_DROP support for the DQO RDA queue format.
> 
> The infrastructure changes of note include an allocation path refactor
> for XDP queues, and a unification of RX buffer sizes across queue
> formats.
> 
> [...]

Here is the summary with links:
  - [net-next,1/6] gve: remove xdp_xsk_done and xdp_xsk_wakeup statistics
    https://git.kernel.org/netdev/net-next/c/c2b900958535
  - [net-next,2/6] gve: introduce config-based allocation for XDP
    https://git.kernel.org/netdev/net-next/c/542a58f1b090
  - [net-next,3/6] gve: update GQ RX to use buf_size
    https://git.kernel.org/netdev/net-next/c/57a070c2672b
  - [net-next,4/6] gve: merge packet buffer size fields
    https://git.kernel.org/netdev/net-next/c/904effd02df7
  - [net-next,5/6] gve: update XDP allocation path support RX buffer posting
    https://git.kernel.org/netdev/net-next/c/346fb86ddd86
  - [net-next,6/6] gve: add XDP DROP and PASS support for DQ
    https://git.kernel.org/netdev/net-next/c/293b49361f91

You are awesome, thank you!
-- 
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html