[PATCH net-next v5 00/24][pull request] Add support for providers with large rx buffer

[PATCH net-next v5 00/24][pull request] Add support for providers with large rx buffer

[Pavel Begunkov]

Many modern network cards support configurable rx buffer lengths larger
than typically used PAGE_SIZE. When paired with hw-gro larger rx buffer
sizes can drastically reduce the number of buffers traversing the stack
and save a lot of processing time. Another benefit for memory providers
like zcrx is that the userspace will be getting larger contiguous chunks
as well.

This series adds net infrastructure for memory providers configuring
the size and implements it for bnxt. It'll be used by io_uring/zcrx,
which is intentionally separated to simplify merging. You can find
a branch that includes zcrx changes at [1] and an example liburing
program at [3].

It's an opt-in feature for drivers, they should advertise support for
the parameter in the qops and must check if the hardware supports
the given size. It's limited to memory providers as it drastically
simplifies the series comparing with previous revisions and detangles
it from ethtool api discussions.

The idea was first floated around by Saeed during netdev conf 2024.
The series also borrows some patches from [2].

Benchmarks with zcrx [3] show up to ~30% improvement in CPU util.
E.g. comparison for 4K vs 32K buffers using a 200Gbit NIC, napi and
userspace pinned to the same CPU:

packets=23987040 (MB=2745098), rps=199559 (MB/s=22837)
CPU    %usr   %nice    %sys %iowait    %irq   %soft   %idle
  0    1.53    0.00   27.78    2.72    1.31   66.45    0.22
packets=24078368 (MB=2755550), rps=200319 (MB/s=22924)
CPU    %usr   %nice    %sys %iowait    %irq   %soft   %idle
  0    0.69    0.00    8.26   31.65    1.83   57.00    0.57

netdev + zcrx changes:
[1] https://github.com/isilence/linux.git zcrx/large-buffers-v5
Per queue configuration series:
[2] https://lore.kernel.org/all/20250421222827.283737-1-kuba@kernel.org/
Liburing example:
[3] https://github.com/isilence/liburing.git zcrx/rx-buf-len

---
The following changes since commit 3a8660878839faadb4f1a6dd72c3179c1df56787:

  Linux 6.18-rc1 (2025-10-12 13:42:36 -0700)

are available in the Git repository at:

  https://github.com/isilence/linux.git tags/net-for-6.19-queue-rx-buf-len-v5

for you to fetch changes up to f389276330412ec4305fb423944261e78490f06a:

  eth: bnxt: allow providers to set rx buf size (2025-10-14 00:11:59 +0100)


v5: Remove all unnecessary bits like configuration via netlink, and
    multi-stage queue configuration.

v4: https://lore.kernel.org/all/cover.1760364551.git.asml.silence@gmail.com/
    - Update fbnic qops
    - Propagate max buf len for hns3
    - Use configured buf size in __bnxt_alloc_rx_netmem
    - Minor stylistic changes
v3: https://lore.kernel.org/all/cover.1755499375.git.asml.silence@gmail.com/
    - Rebased, excluded zcrx specific patches
    - Set agg_size_fac to 1 on warning
v2: https://lore.kernel.org/all/cover.1754657711.git.asml.silence@gmail.com/
    - Add MAX_PAGE_ORDER check on pp init
    - Applied comments rewording
    - Adjust pp.max_len based on order
    - Patch up mlx5 queue callbacks after rebase
    - Minor ->queue_mgmt_ops refactoring
    - Rebased to account for both fill level and agg_size_fac
    - Pass providers buf length in struct pp_memory_provider_params and
      apply it in __netdev_queue_confi().
    - Use ->supported_ring_params to validate drivers support of set
      qcfg parameters.

Jakub Kicinski (1):
  eth: bnxt: adjust the fill level of agg queues with larger buffers

Pavel Begunkov (5):
  net: page_pool: sanitise allocation order
  net: memzero mp params when closing a queue
  net: let pp memory provider to specify rx buf len
  eth: bnxt: store rx buffer size per queue
  eth: bnxt: allow providers to set rx buf size

 drivers/net/ethernet/broadcom/bnxt/bnxt.c     | 118 ++++++++++++++----
 drivers/net/ethernet/broadcom/bnxt/bnxt.h     |   2 +
 drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.c |   6 +-
 drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.h |   2 +-
 include/net/netdev_queues.h                   |   9 ++
 include/net/page_pool/types.h                 |   1 +
 net/core/netdev_rx_queue.c                    |  14 ++-
 net/core/page_pool.c                          |   3 +
 8 files changed, 118 insertions(+), 37 deletions(-)

-- 
2.49.0

[PATCH net-next v5 1/6] net: page_pool: sanitise allocation order

[Pavel Begunkov]

We're going to give more control over rx buffer sizes to user space, and
since we can't always rely on driver validation, let's sanitise it in
page_pool_init() as well. Note that we only need to reject over
MAX_PAGE_ORDER allocations for normal page pools, as current memory
providers don't need to use the buddy allocator and must check the order
on init.i

Suggested-by: Stanislav Fomichev <stfomichev@gmail.com>
Reviewed-by: Mina Almasry <almasrymina@google.com>
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
---
 net/core/page_pool.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/net/core/page_pool.c b/net/core/page_pool.c
index 1a5edec485f1..635c77e8050b 100644
--- a/net/core/page_pool.c
+++ b/net/core/page_pool.c
@@ -301,6 +301,9 @@ static int page_pool_init(struct page_pool *pool,
 		}
 
 		static_branch_inc(&page_pool_mem_providers);
+	} else if (pool->p.order > MAX_PAGE_ORDER) {
+		err = -EINVAL;
+		goto free_ptr_ring;
 	}
 
 	return 0;
-- 
2.49.0

[PATCH net-next v5 2/6] net: memzero mp params when closing a queue

[Pavel Begunkov]

Instead of resetting memory provider parameters one by one in
__net_mp_{open,close}_rxq, memzero the entire structure. It'll be used
to extend the structure.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
---
 net/core/netdev_rx_queue.c | 10 ++++------
 1 file changed, 4 insertions(+), 6 deletions(-)

diff --git a/net/core/netdev_rx_queue.c b/net/core/netdev_rx_queue.c
index c7d9341b7630..a0083f176a9c 100644
--- a/net/core/netdev_rx_queue.c
+++ b/net/core/netdev_rx_queue.c
@@ -139,10 +139,9 @@ int __net_mp_open_rxq(struct net_device *dev, unsigned int rxq_idx,
 
 	rxq->mp_params = *p;
 	ret = netdev_rx_queue_restart(dev, rxq_idx);
-	if (ret) {
-		rxq->mp_params.mp_ops = NULL;
-		rxq->mp_params.mp_priv = NULL;
-	}
+	if (ret)
+		memset(&rxq->mp_params, 0, sizeof(rxq->mp_params));
+
 	return ret;
 }
 
@@ -179,8 +178,7 @@ void __net_mp_close_rxq(struct net_device *dev, unsigned int ifq_idx,
 			 rxq->mp_params.mp_priv != old_p->mp_priv))
 		return;
 
-	rxq->mp_params.mp_ops = NULL;
-	rxq->mp_params.mp_priv = NULL;
+	memset(&rxq->mp_params, 0, sizeof(rxq->mp_params));
 	err = netdev_rx_queue_restart(dev, ifq_idx);
 	WARN_ON(err && err != -ENETDOWN);
 }
-- 
2.49.0

[PATCH net-next v5 3/6] net: let pp memory provider to specify rx buf len

[Pavel Begunkov]

Allow memory providers to configure rx queues with a specific receive
buffer length. Pass it in struct pp_memory_provider_params, which is
copied into the queue, so it's preserved across queue restarts. It's an
opt-in feature for drivers, which they can enable by setting
NDO_QUEUE_RX_BUF_SIZE to their struct netdev_queue_mgmt_ops.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
---
 include/net/netdev_queues.h   | 9 +++++++++
 include/net/page_pool/types.h | 1 +
 net/core/netdev_rx_queue.c    | 4 ++++
 3 files changed, 14 insertions(+)

diff --git a/include/net/netdev_queues.h b/include/net/netdev_queues.h
index cd00e0406cf4..2e6bcec1e1e3 100644
--- a/include/net/netdev_queues.h
+++ b/include/net/netdev_queues.h
@@ -111,6 +111,11 @@ void netdev_stat_queue_sum(struct net_device *netdev,
 			   int tx_start, int tx_end,
 			   struct netdev_queue_stats_tx *tx_sum);
 
+enum {
+	/* queue restart support custom rx buffer sizes */
+	NDO_QUEUE_RX_BUF_SIZE		= 0x1,
+};
+
 /**
  * struct netdev_queue_mgmt_ops - netdev ops for queue management
  *
@@ -130,6 +135,8 @@ void netdev_stat_queue_sum(struct net_device *netdev,
  * @ndo_queue_get_dma_dev: Get dma device for zero-copy operations to be used
  *			   for this queue. Return NULL on error.
  *
+ * @supported_params: bitmask of supported features, see NDO_QUEUE_*
+ *
  * Note that @ndo_queue_mem_alloc and @ndo_queue_mem_free may be called while
  * the interface is closed. @ndo_queue_start and @ndo_queue_stop will only
  * be called for an interface which is open.
@@ -149,6 +156,8 @@ struct netdev_queue_mgmt_ops {
 						  int idx);
 	struct device *		(*ndo_queue_get_dma_dev)(struct net_device *dev,
 							 int idx);
+
+	unsigned supported_params;
 };
 
 bool netif_rxq_has_unreadable_mp(struct net_device *dev, int idx);
diff --git a/include/net/page_pool/types.h b/include/net/page_pool/types.h
index 1509a536cb85..be74e4aec7b5 100644
--- a/include/net/page_pool/types.h
+++ b/include/net/page_pool/types.h
@@ -161,6 +161,7 @@ struct memory_provider_ops;
 struct pp_memory_provider_params {
 	void *mp_priv;
 	const struct memory_provider_ops *mp_ops;
+	u32 rx_buf_len;
 };
 
 struct page_pool {
diff --git a/net/core/netdev_rx_queue.c b/net/core/netdev_rx_queue.c
index a0083f176a9c..09d6f97e910e 100644
--- a/net/core/netdev_rx_queue.c
+++ b/net/core/netdev_rx_queue.c
@@ -29,6 +29,10 @@ int netdev_rx_queue_restart(struct net_device *dev, unsigned int rxq_idx)
 	    !qops->ndo_queue_mem_alloc || !qops->ndo_queue_start)
 		return -EOPNOTSUPP;
 
+	if (!(qops->supported_params & NDO_QUEUE_RX_BUF_SIZE) &&
+	    rxq->mp_params.rx_buf_len)
+		return -EOPNOTSUPP;
+
 	netdev_assert_locked(dev);
 
 	new_mem = kvzalloc(qops->ndo_queue_mem_size, GFP_KERNEL);
-- 
2.49.0

[PATCH net-next v5 4/6] eth: bnxt: store rx buffer size per queue

[Pavel Begunkov]

Instead of using a constant buffer length, allow configuring the size
for each queue separately. There is no way to change the length yet, and
it'll be passed from memory providers in a later patch.

Suggested-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
---
 drivers/net/ethernet/broadcom/bnxt/bnxt.c     | 56 +++++++++++--------
 drivers/net/ethernet/broadcom/bnxt/bnxt.h     |  1 +
 drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.c |  6 +-
 drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.h |  2 +-
 4 files changed, 38 insertions(+), 27 deletions(-)

diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.c b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
index 3fc33b1b4dfb..61943de13774 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
@@ -905,7 +905,7 @@ static void bnxt_tx_int(struct bnxt *bp, struct bnxt_napi *bnapi, int budget)
 
 static bool bnxt_separate_head_pool(struct bnxt_rx_ring_info *rxr)
 {
-	return rxr->need_head_pool || PAGE_SIZE > BNXT_RX_PAGE_SIZE;
+	return rxr->need_head_pool || rxr->rx_page_size < PAGE_SIZE;
 }
 
 static struct page *__bnxt_alloc_rx_page(struct bnxt *bp, dma_addr_t *mapping,
@@ -915,9 +915,9 @@ static struct page *__bnxt_alloc_rx_page(struct bnxt *bp, dma_addr_t *mapping,
 {
 	struct page *page;
 
-	if (PAGE_SIZE > BNXT_RX_PAGE_SIZE) {
+	if (rxr->rx_page_size < PAGE_SIZE) {
 		page = page_pool_dev_alloc_frag(rxr->page_pool, offset,
-						BNXT_RX_PAGE_SIZE);
+						rxr->rx_page_size);
 	} else {
 		page = page_pool_dev_alloc_pages(rxr->page_pool);
 		*offset = 0;
@@ -936,8 +936,9 @@ static netmem_ref __bnxt_alloc_rx_netmem(struct bnxt *bp, dma_addr_t *mapping,
 {
 	netmem_ref netmem;
 
-	if (PAGE_SIZE > BNXT_RX_PAGE_SIZE) {
-		netmem = page_pool_alloc_frag_netmem(rxr->page_pool, offset, BNXT_RX_PAGE_SIZE, gfp);
+	if (rxr->rx_page_size < PAGE_SIZE) {
+		netmem = page_pool_alloc_frag_netmem(rxr->page_pool, offset,
+						     rxr->rx_page_size, gfp);
 	} else {
 		netmem = page_pool_alloc_netmems(rxr->page_pool, gfp);
 		*offset = 0;
@@ -1155,9 +1156,9 @@ static struct sk_buff *bnxt_rx_multi_page_skb(struct bnxt *bp,
 		return NULL;
 	}
 	dma_addr -= bp->rx_dma_offset;
-	dma_sync_single_for_cpu(&bp->pdev->dev, dma_addr, BNXT_RX_PAGE_SIZE,
+	dma_sync_single_for_cpu(&bp->pdev->dev, dma_addr, rxr->rx_page_size,
 				bp->rx_dir);
-	skb = napi_build_skb(data_ptr - bp->rx_offset, BNXT_RX_PAGE_SIZE);
+	skb = napi_build_skb(data_ptr - bp->rx_offset, rxr->rx_page_size);
 	if (!skb) {
 		page_pool_recycle_direct(rxr->page_pool, page);
 		return NULL;
@@ -1189,7 +1190,7 @@ static struct sk_buff *bnxt_rx_page_skb(struct bnxt *bp,
 		return NULL;
 	}
 	dma_addr -= bp->rx_dma_offset;
-	dma_sync_single_for_cpu(&bp->pdev->dev, dma_addr, BNXT_RX_PAGE_SIZE,
+	dma_sync_single_for_cpu(&bp->pdev->dev, dma_addr, rxr->rx_page_size,
 				bp->rx_dir);
 
 	if (unlikely(!payload))
@@ -1203,7 +1204,7 @@ static struct sk_buff *bnxt_rx_page_skb(struct bnxt *bp,
 
 	skb_mark_for_recycle(skb);
 	off = (void *)data_ptr - page_address(page);
-	skb_add_rx_frag(skb, 0, page, off, len, BNXT_RX_PAGE_SIZE);
+	skb_add_rx_frag(skb, 0, page, off, len, rxr->rx_page_size);
 	memcpy(skb->data - NET_IP_ALIGN, data_ptr - NET_IP_ALIGN,
 	       payload + NET_IP_ALIGN);
 
@@ -1288,7 +1289,7 @@ static u32 __bnxt_rx_agg_netmems(struct bnxt *bp,
 		if (skb) {
 			skb_add_rx_frag_netmem(skb, i, cons_rx_buf->netmem,
 					       cons_rx_buf->offset,
-					       frag_len, BNXT_RX_PAGE_SIZE);
+					       frag_len, rxr->rx_page_size);
 		} else {
 			skb_frag_t *frag = &shinfo->frags[i];
 
@@ -1313,7 +1314,7 @@ static u32 __bnxt_rx_agg_netmems(struct bnxt *bp,
 			if (skb) {
 				skb->len -= frag_len;
 				skb->data_len -= frag_len;
-				skb->truesize -= BNXT_RX_PAGE_SIZE;
+				skb->truesize -= rxr->rx_page_size;
 			}
 
 			--shinfo->nr_frags;
@@ -1328,7 +1329,7 @@ static u32 __bnxt_rx_agg_netmems(struct bnxt *bp,
 		}
 
 		page_pool_dma_sync_netmem_for_cpu(rxr->page_pool, netmem, 0,
-						  BNXT_RX_PAGE_SIZE);
+						  rxr->rx_page_size);
 
 		total_frag_len += frag_len;
 		prod = NEXT_RX_AGG(prod);
@@ -2281,8 +2282,7 @@ static int bnxt_rx_pkt(struct bnxt *bp, struct bnxt_cp_ring_info *cpr,
 			if (!skb)
 				goto oom_next_rx;
 		} else {
-			skb = bnxt_xdp_build_skb(bp, skb, agg_bufs,
-						 rxr->page_pool, &xdp);
+			skb = bnxt_xdp_build_skb(bp, skb, agg_bufs, rxr, &xdp);
 			if (!skb) {
 				/* we should be able to free the old skb here */
 				bnxt_xdp_buff_frags_free(rxr, &xdp);
@@ -3828,11 +3828,13 @@ static int bnxt_alloc_rx_page_pool(struct bnxt *bp,
 	pp.pool_size = bp->rx_agg_ring_size / agg_size_fac;
 	if (BNXT_RX_PAGE_MODE(bp))
 		pp.pool_size += bp->rx_ring_size / rx_size_fac;
+
+	pp.order = get_order(rxr->rx_page_size);
 	pp.nid = numa_node;
 	pp.netdev = bp->dev;
 	pp.dev = &bp->pdev->dev;
 	pp.dma_dir = bp->rx_dir;
-	pp.max_len = PAGE_SIZE;
+	pp.max_len = PAGE_SIZE << pp.order;
 	pp.flags = PP_FLAG_DMA_MAP | PP_FLAG_DMA_SYNC_DEV |
 		   PP_FLAG_ALLOW_UNREADABLE_NETMEM;
 	pp.queue_idx = rxr->bnapi->index;
@@ -3843,7 +3845,10 @@ static int bnxt_alloc_rx_page_pool(struct bnxt *bp,
 	rxr->page_pool = pool;
 
 	rxr->need_head_pool = page_pool_is_unreadable(pool);
+	rxr->need_head_pool |= !!pp.order;
 	if (bnxt_separate_head_pool(rxr)) {
+		pp.order = 0;
+		pp.max_len = PAGE_SIZE;
 		pp.pool_size = min(bp->rx_ring_size / rx_size_fac, 1024);
 		pp.flags = PP_FLAG_DMA_MAP | PP_FLAG_DMA_SYNC_DEV;
 		pool = page_pool_create(&pp);
@@ -4319,6 +4324,8 @@ static void bnxt_init_ring_struct(struct bnxt *bp)
 		if (!rxr)
 			goto skip_rx;
 
+		rxr->rx_page_size = BNXT_RX_PAGE_SIZE;
+
 		ring = &rxr->rx_ring_struct;
 		rmem = &ring->ring_mem;
 		rmem->nr_pages = bp->rx_nr_pages;
@@ -4478,7 +4485,7 @@ static void bnxt_init_one_rx_agg_ring_rxbd(struct bnxt *bp,
 	ring = &rxr->rx_agg_ring_struct;
 	ring->fw_ring_id = INVALID_HW_RING_ID;
 	if ((bp->flags & BNXT_FLAG_AGG_RINGS)) {
-		type = ((u32)BNXT_RX_PAGE_SIZE << RX_BD_LEN_SHIFT) |
+		type = ((u32)rxr->rx_page_size << RX_BD_LEN_SHIFT) |
 			RX_BD_TYPE_RX_AGG_BD | RX_BD_FLAGS_SOP;
 
 		bnxt_init_rxbd_pages(ring, type);
@@ -7045,6 +7052,7 @@ static void bnxt_hwrm_ring_grp_free(struct bnxt *bp)
 
 static void bnxt_set_rx_ring_params_p5(struct bnxt *bp, u32 ring_type,
 				       struct hwrm_ring_alloc_input *req,
+				       struct bnxt_rx_ring_info *rxr,
 				       struct bnxt_ring_struct *ring)
 {
 	struct bnxt_ring_grp_info *grp_info = &bp->grp_info[ring->grp_idx];
@@ -7054,7 +7062,7 @@ static void bnxt_set_rx_ring_params_p5(struct bnxt *bp, u32 ring_type,
 	if (ring_type == HWRM_RING_ALLOC_AGG) {
 		req->ring_type = RING_ALLOC_REQ_RING_TYPE_RX_AGG;
 		req->rx_ring_id = cpu_to_le16(grp_info->rx_fw_ring_id);
-		req->rx_buf_size = cpu_to_le16(BNXT_RX_PAGE_SIZE);
+		req->rx_buf_size = cpu_to_le16(rxr->rx_page_size);
 		enables |= RING_ALLOC_REQ_ENABLES_RX_RING_ID_VALID;
 	} else {
 		req->rx_buf_size = cpu_to_le16(bp->rx_buf_use_size);
@@ -7068,6 +7076,7 @@ static void bnxt_set_rx_ring_params_p5(struct bnxt *bp, u32 ring_type,
 }
 
 static int hwrm_ring_alloc_send_msg(struct bnxt *bp,
+				    struct bnxt_rx_ring_info *rxr,
 				    struct bnxt_ring_struct *ring,
 				    u32 ring_type, u32 map_index)
 {
@@ -7124,7 +7133,8 @@ static int hwrm_ring_alloc_send_msg(struct bnxt *bp,
 			      cpu_to_le32(bp->rx_ring_mask + 1) :
 			      cpu_to_le32(bp->rx_agg_ring_mask + 1);
 		if (bp->flags & BNXT_FLAG_CHIP_P5_PLUS)
-			bnxt_set_rx_ring_params_p5(bp, ring_type, req, ring);
+			bnxt_set_rx_ring_params_p5(bp, ring_type, req,
+						   rxr, ring);
 		break;
 	case HWRM_RING_ALLOC_CMPL:
 		req->ring_type = RING_ALLOC_REQ_RING_TYPE_L2_CMPL;
@@ -7272,7 +7282,7 @@ static int bnxt_hwrm_rx_ring_alloc(struct bnxt *bp,
 	u32 map_idx = bnapi->index;
 	int rc;
 
-	rc = hwrm_ring_alloc_send_msg(bp, ring, type, map_idx);
+	rc = hwrm_ring_alloc_send_msg(bp, rxr, ring, type, map_idx);
 	if (rc)
 		return rc;
 
@@ -7292,7 +7302,7 @@ static int bnxt_hwrm_rx_agg_ring_alloc(struct bnxt *bp,
 	int rc;
 
 	map_idx = grp_idx + bp->rx_nr_rings;
-	rc = hwrm_ring_alloc_send_msg(bp, ring, type, map_idx);
+	rc = hwrm_ring_alloc_send_msg(bp, rxr, ring, type, map_idx);
 	if (rc)
 		return rc;
 
@@ -7316,7 +7326,7 @@ static int bnxt_hwrm_cp_ring_alloc_p5(struct bnxt *bp,
 
 	ring = &cpr->cp_ring_struct;
 	ring->handle = BNXT_SET_NQ_HDL(cpr);
-	rc = hwrm_ring_alloc_send_msg(bp, ring, type, map_idx);
+	rc = hwrm_ring_alloc_send_msg(bp, NULL, ring, type, map_idx);
 	if (rc)
 		return rc;
 	bnxt_set_db(bp, &cpr->cp_db, type, map_idx, ring->fw_ring_id);
@@ -7331,7 +7341,7 @@ static int bnxt_hwrm_tx_ring_alloc(struct bnxt *bp,
 	const u32 type = HWRM_RING_ALLOC_TX;
 	int rc;
 
-	rc = hwrm_ring_alloc_send_msg(bp, ring, type, tx_idx);
+	rc = hwrm_ring_alloc_send_msg(bp, NULL, ring, type, tx_idx);
 	if (rc)
 		return rc;
 	bnxt_set_db(bp, &txr->tx_db, type, tx_idx, ring->fw_ring_id);
@@ -7357,7 +7367,7 @@ static int bnxt_hwrm_ring_alloc(struct bnxt *bp)
 
 		vector = bp->irq_tbl[map_idx].vector;
 		disable_irq_nosync(vector);
-		rc = hwrm_ring_alloc_send_msg(bp, ring, type, map_idx);
+		rc = hwrm_ring_alloc_send_msg(bp, NULL, ring, type, map_idx);
 		if (rc) {
 			enable_irq(vector);
 			goto err_out;
diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.h b/drivers/net/ethernet/broadcom/bnxt/bnxt.h
index 741b2d854789..3a53d9073c4d 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt.h
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.h
@@ -1106,6 +1106,7 @@ struct bnxt_rx_ring_info {
 
 	unsigned long		*rx_agg_bmap;
 	u16			rx_agg_bmap_size;
+	u16			rx_page_size;
 	bool                    need_head_pool;
 
 	dma_addr_t		rx_desc_mapping[MAX_RX_PAGES];
diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.c b/drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.c
index 3e77a96e5a3e..619235b151a4 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.c
@@ -183,7 +183,7 @@ void bnxt_xdp_buff_init(struct bnxt *bp, struct bnxt_rx_ring_info *rxr,
 			u16 cons, u8 *data_ptr, unsigned int len,
 			struct xdp_buff *xdp)
 {
-	u32 buflen = BNXT_RX_PAGE_SIZE;
+	u32 buflen = rxr->rx_page_size;
 	struct bnxt_sw_rx_bd *rx_buf;
 	struct pci_dev *pdev;
 	dma_addr_t mapping;
@@ -461,7 +461,7 @@ int bnxt_xdp(struct net_device *dev, struct netdev_bpf *xdp)
 
 struct sk_buff *
 bnxt_xdp_build_skb(struct bnxt *bp, struct sk_buff *skb, u8 num_frags,
-		   struct page_pool *pool, struct xdp_buff *xdp)
+		   struct bnxt_rx_ring_info *rxr, struct xdp_buff *xdp)
 {
 	struct skb_shared_info *sinfo = xdp_get_shared_info_from_buff(xdp);
 
@@ -469,7 +469,7 @@ bnxt_xdp_build_skb(struct bnxt *bp, struct sk_buff *skb, u8 num_frags,
 		return NULL;
 
 	xdp_update_skb_frags_info(skb, num_frags, sinfo->xdp_frags_size,
-				  BNXT_RX_PAGE_SIZE * num_frags,
+				  rxr->rx_page_size * num_frags,
 				  xdp_buff_get_skb_flags(xdp));
 	return skb;
 }
diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.h b/drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.h
index 220285e190fc..8933a0dec09a 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.h
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.h
@@ -32,6 +32,6 @@ void bnxt_xdp_buff_init(struct bnxt *bp, struct bnxt_rx_ring_info *rxr,
 void bnxt_xdp_buff_frags_free(struct bnxt_rx_ring_info *rxr,
 			      struct xdp_buff *xdp);
 struct sk_buff *bnxt_xdp_build_skb(struct bnxt *bp, struct sk_buff *skb,
-				   u8 num_frags, struct page_pool *pool,
+				   u8 num_frags, struct bnxt_rx_ring_info *rxr,
 				   struct xdp_buff *xdp);
 #endif
-- 
2.49.0

[PATCH net-next v5 5/6] eth: bnxt: adjust the fill level of agg queues with larger buffers

[Pavel Begunkov]

From: Jakub Kicinski <kuba@kernel.org>

The driver tries to provision more agg buffers than header buffers
since multiple agg segments can reuse the same header. The calculation
/ heuristic tries to provide enough pages for 65k of data for each header
(or 4 frags per header if the result is too big). This calculation is
currently global to the adapter. If we increase the buffer sizes 8x
we don't want 8x the amount of memory sitting on the rings.
Luckily we don't have to fill the rings completely, adjust
the fill level dynamically in case particular queue has buffers
larger than the global size.

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
[pavel: rebase on top of agg_size_fac, assert agg_size_fac]
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
---
 drivers/net/ethernet/broadcom/bnxt/bnxt.c | 28 +++++++++++++++++++----
 1 file changed, 24 insertions(+), 4 deletions(-)

diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.c b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
index 61943de13774..950e60d204cb 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
@@ -3816,16 +3816,34 @@ static void bnxt_free_rx_rings(struct bnxt *bp)
 	}
 }
 
+static int bnxt_rx_agg_ring_fill_level(struct bnxt *bp,
+				       struct bnxt_rx_ring_info *rxr)
+{
+	/* User may have chosen larger than default rx_page_size,
+	 * we keep the ring sizes uniform and also want uniform amount
+	 * of bytes consumed per ring, so cap how much of the rings we fill.
+	 */
+	int fill_level = bp->rx_agg_ring_size;
+
+	if (rxr->rx_page_size > BNXT_RX_PAGE_SIZE)
+		fill_level /= rxr->rx_page_size / BNXT_RX_PAGE_SIZE;
+
+	return fill_level;
+}
+
 static int bnxt_alloc_rx_page_pool(struct bnxt *bp,
 				   struct bnxt_rx_ring_info *rxr,
 				   int numa_node)
 {
-	const unsigned int agg_size_fac = PAGE_SIZE / BNXT_RX_PAGE_SIZE;
+	unsigned int agg_size_fac = rxr->rx_page_size / BNXT_RX_PAGE_SIZE;
 	const unsigned int rx_size_fac = PAGE_SIZE / SZ_4K;
 	struct page_pool_params pp = { 0 };
 	struct page_pool *pool;
 
-	pp.pool_size = bp->rx_agg_ring_size / agg_size_fac;
+	if (WARN_ON_ONCE(agg_size_fac == 0))
+		agg_size_fac = 1;
+
+	pp.pool_size = bnxt_rx_agg_ring_fill_level(bp, rxr) / agg_size_fac;
 	if (BNXT_RX_PAGE_MODE(bp))
 		pp.pool_size += bp->rx_ring_size / rx_size_fac;
 
@@ -4403,11 +4421,13 @@ static void bnxt_alloc_one_rx_ring_netmem(struct bnxt *bp,
 					  struct bnxt_rx_ring_info *rxr,
 					  int ring_nr)
 {
+	int fill_level, i;
 	u32 prod;
-	int i;
+
+	fill_level = bnxt_rx_agg_ring_fill_level(bp, rxr);
 
 	prod = rxr->rx_agg_prod;
-	for (i = 0; i < bp->rx_agg_ring_size; i++) {
+	for (i = 0; i < fill_level; i++) {
 		if (bnxt_alloc_rx_netmem(bp, rxr, prod, GFP_KERNEL)) {
 			netdev_warn(bp->dev, "init'ed rx ring %d with %d/%d pages only\n",
 				    ring_nr, i, bp->rx_agg_ring_size);
-- 
2.49.0

[PATCH net-next v5 6/6] eth: bnxt: allow providers to set rx buf size

[Pavel Begunkov]

Implement NDO_QUEUE_RX_BUF_SIZE and take the rx buf size from the memory
providers.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
---
 drivers/net/ethernet/broadcom/bnxt/bnxt.c | 34 +++++++++++++++++++++++
 drivers/net/ethernet/broadcom/bnxt/bnxt.h |  1 +
 2 files changed, 35 insertions(+)

diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.c b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
index 950e60d204cb..8d0e69dc900e 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
@@ -15906,16 +15906,46 @@ static const struct netdev_stat_ops bnxt_stat_ops = {
 	.get_base_stats		= bnxt_get_base_stats,
 };
 
+static ssize_t bnxt_get_rx_buf_size(struct bnxt *bp, int rxq_idx)
+{
+	struct netdev_rx_queue *rxq = __netif_get_rx_queue(bp->dev, rxq_idx);
+	size_t rx_buf_size;
+
+	rx_buf_size = rxq->mp_params.rx_buf_len;
+	if (!rx_buf_size)
+		return BNXT_RX_PAGE_SIZE;
+
+	/* Older chips need MSS calc so rx_buf_len is not supported,
+	 * but we don't set queue ops for them so we should never get here.
+	 */
+	if (!(bp->flags & BNXT_FLAG_CHIP_P5_PLUS))
+		return -EINVAL;
+
+	if (!is_power_of_2(rx_buf_size))
+		return -ERANGE;
+
+	if (rx_buf_size < BNXT_RX_PAGE_SIZE ||
+	    rx_buf_size > BNXT_MAX_RX_PAGE_SIZE)
+		return -ERANGE;
+
+	return rx_buf_size;
+}
+
 static int bnxt_queue_mem_alloc(struct net_device *dev, void *qmem, int idx)
 {
 	struct bnxt_rx_ring_info *rxr, *clone;
 	struct bnxt *bp = netdev_priv(dev);
 	struct bnxt_ring_struct *ring;
+	ssize_t rx_buf_size;
 	int rc;
 
 	if (!bp->rx_ring)
 		return -ENETDOWN;
 
+	rx_buf_size = bnxt_get_rx_buf_size(bp, idx);
+	if (rx_buf_size < 0)
+		return rx_buf_size;
+
 	rxr = &bp->rx_ring[idx];
 	clone = qmem;
 	memcpy(clone, rxr, sizeof(*rxr));
@@ -15927,6 +15957,7 @@ static int bnxt_queue_mem_alloc(struct net_device *dev, void *qmem, int idx)
 	clone->rx_sw_agg_prod = 0;
 	clone->rx_next_cons = 0;
 	clone->need_head_pool = false;
+	clone->rx_page_size = rx_buf_size;
 
 	rc = bnxt_alloc_rx_page_pool(bp, clone, rxr->page_pool->p.nid);
 	if (rc)
@@ -16053,6 +16084,8 @@ static void bnxt_copy_rx_ring(struct bnxt *bp,
 	src_ring = &src->rx_agg_ring_struct;
 	src_rmem = &src_ring->ring_mem;
 
+	dst->rx_page_size = src->rx_page_size;
+
 	WARN_ON(dst_rmem->nr_pages != src_rmem->nr_pages);
 	WARN_ON(dst_rmem->page_size != src_rmem->page_size);
 	WARN_ON(dst_rmem->flags != src_rmem->flags);
@@ -16205,6 +16238,7 @@ static const struct netdev_queue_mgmt_ops bnxt_queue_mgmt_ops = {
 	.ndo_queue_mem_free	= bnxt_queue_mem_free,
 	.ndo_queue_start	= bnxt_queue_start,
 	.ndo_queue_stop		= bnxt_queue_stop,
+	.supported_params	= NDO_QUEUE_RX_BUF_SIZE,
 };
 
 static void bnxt_remove_one(struct pci_dev *pdev)
diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.h b/drivers/net/ethernet/broadcom/bnxt/bnxt.h
index 3a53d9073c4d..d84c0de231b3 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt.h
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.h
@@ -759,6 +759,7 @@ struct nqe_cn {
 #endif
 
 #define BNXT_RX_PAGE_SIZE (1 << BNXT_RX_PAGE_SHIFT)
+#define BNXT_MAX_RX_PAGE_SIZE (1 << 15)
 
 #define BNXT_MAX_MTU		9500
 
-- 
2.49.0

Re: [PATCH net-next v5 00/24][pull request] Add support for providers with large rx buffer

[Pavel Begunkov]

Oops, should be 0/6 in the subject.

On 10/14/25 14:01, Pavel Begunkov wrote:
> Many modern network cards support configurable rx buffer lengths larger
> than typically used PAGE_SIZE. When paired with hw-gro larger rx buffer
> sizes can drastically reduce the number of buffers traversing the stack
> and save a lot of processing time. Another benefit for memory providers
> like zcrx is that the userspace will be getting larger contiguous chunks
> as well.
> 
> This series adds net infrastructure for memory providers configuring
> the size and implements it for bnxt. It'll be used by io_uring/zcrx,
> which is intentionally separated to simplify merging. You can find
> a branch that includes zcrx changes at [1] and an example liburing
> program at [3].

-- 
Pavel Begunkov