[PATCH net 0/4] eth: fbnic: fix and improve header/data split configuration

[PATCH net 0/4] eth: fbnic: fix and improve header/data split configuration

[Bobby Eshleman]

This series fixes TCP HDS configuration in the fbnic driver and adds a
devmem selftest for the intended behavior.

The issues fixed include setting the correct CSR for EN_HDR_SPLIT,
adjusting the hds threshold clamp to avoid long headers overflowing into
the payload page, and configuring the device to use L4/L3/L2 header
boundaries when present by programming the DMA hint bit unconditionally
for all steering rule types.

Prior to these fixes, small payloads fail devmem.check_rx_hds().

Testing:

IFC0=enp1s0
IFC1=eth0

export REMOTE_TYPE=netns
export REMOTE_ARGS=ns-remote
export NETIF=$IFC0
export LOCAL_V4=192.0.3.1
export REMOTE_V4=192.0.3.2
sysctl -w net.ipv6.conf.$IFC0.keep_addr_on_down=1

./tools/testing/selftests/drivers/net/hw/devmem.py

TAP version 13
1..4
ok 1 devmem.check_rx
ok 2 devmem.check_tx
ok 3 devmem.check_tx_chunks
ok 4 devmem.check_rx_hds

To: Alexander Duyck <alexanderduyck@fb.com>
To: Jakub Kicinski <kuba@kernel.org>
To: kernel-team@meta.com
To: Andrew Lunn <andrew+netdev@lunn.ch>
To: David S. Miller <davem@davemloft.net>
To: Eric Dumazet <edumazet@google.com>
To: Paolo Abeni <pabeni@redhat.com>
To: Mohsin Bashir <mohsin.bashr@gmail.com>
To: Shuah Khan <shuah@kernel.org>
Cc: netdev@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Cc: linux-kselftest@vger.kernel.org

Signed-off-by: Bobby Eshleman <bobbyeshleman@meta.com>
---
Bobby Eshleman (4):
      eth: fbnic: set FBNIC_QUEUE_RDE_CTL0_EN_HDR_SPLIT on RDE_CTL0
      eth: fbnic: increase FBNIC_HDR_BYTES_MIN from 128 to 256 bytes
      eth: fbnic: set DMA_HINT_L4 for all flows
      selftests: drv-net: add HDS payload sweep test for devmem TCP

 drivers/net/ethernet/meta/fbnic/fbnic_ethtool.c   |  3 +++
 drivers/net/ethernet/meta/fbnic/fbnic_rpc.c       |  5 ++---
 drivers/net/ethernet/meta/fbnic/fbnic_txrx.c      | 25 +++++++++++++----------
 drivers/net/ethernet/meta/fbnic/fbnic_txrx.h      |  2 +-
 tools/testing/selftests/drivers/net/hw/devmem.py  | 19 ++++++++++++++++-
 tools/testing/selftests/drivers/net/hw/ncdevmem.c | 11 +++++++++-
 6 files changed, 48 insertions(+), 17 deletions(-)
---
base-commit: bf9cf80cab81e39701861a42877a28295ade266f
change-id: 20260211-fbnic-tcp-hds-fixes-e38149aabab8

Best regards,
-- 
Bobby Eshleman <bobbyeshleman@meta.com>

[PATCH net 1/4] eth: fbnic: set FBNIC_QUEUE_RDE_CTL0_EN_HDR_SPLIT on RDE_CTL0

[Bobby Eshleman]

From: Bobby Eshleman <bobbyeshleman@meta.com>

Fix EN_HDR_SPLIT configuration by writing the field to RDE_CTL0 instead
of RDE_CTL1.

Because drop mode configuration and header splitting enablement both use
RDE_CTL0, we consolidate these configurations into the single function
fbnic_config_drop_mode.

Fixes: 2b30fc01a6c7 ("eth: fbnic: Add support for HDS configuration")
Signed-off-by: Bobby Eshleman <bobbyeshleman@meta.com>
Acked-by: Mohsin Bashir <mohsin.bashr@gmail.com>
---
 drivers/net/ethernet/meta/fbnic/fbnic_txrx.c | 25 ++++++++++++++-----------
 1 file changed, 14 insertions(+), 11 deletions(-)

diff --git a/drivers/net/ethernet/meta/fbnic/fbnic_txrx.c b/drivers/net/ethernet/meta/fbnic/fbnic_txrx.c
index 13d508ce637f..e119526fce14 100644
--- a/drivers/net/ethernet/meta/fbnic/fbnic_txrx.c
+++ b/drivers/net/ethernet/meta/fbnic/fbnic_txrx.c
@@ -2575,7 +2575,8 @@ static void fbnic_enable_bdq(struct fbnic_ring *hpq, struct fbnic_ring *ppq)
 }
 
 static void fbnic_config_drop_mode_rcq(struct fbnic_napi_vector *nv,
-				       struct fbnic_ring *rcq, bool tx_pause)
+				       struct fbnic_ring *rcq, bool tx_pause,
+				       bool hdr_split)
 {
 	struct fbnic_net *fbn = netdev_priv(nv->napi.dev);
 	u32 drop_mode, rcq_ctl;
@@ -2588,22 +2589,26 @@ static void fbnic_config_drop_mode_rcq(struct fbnic_napi_vector *nv,
 	/* Specify packet layout */
 	rcq_ctl = FIELD_PREP(FBNIC_QUEUE_RDE_CTL0_DROP_MODE_MASK, drop_mode) |
 	    FIELD_PREP(FBNIC_QUEUE_RDE_CTL0_MIN_HROOM_MASK, FBNIC_RX_HROOM) |
-	    FIELD_PREP(FBNIC_QUEUE_RDE_CTL0_MIN_TROOM_MASK, FBNIC_RX_TROOM);
+	    FIELD_PREP(FBNIC_QUEUE_RDE_CTL0_MIN_TROOM_MASK, FBNIC_RX_TROOM) |
+	    FIELD_PREP(FBNIC_QUEUE_RDE_CTL0_EN_HDR_SPLIT, hdr_split);
 
 	fbnic_ring_wr32(rcq, FBNIC_QUEUE_RDE_CTL0, rcq_ctl);
 }
 
-void fbnic_config_drop_mode(struct fbnic_net *fbn, bool tx_pause)
+void fbnic_config_drop_mode(struct fbnic_net *fbn, bool txp)
 {
+	bool hds;
 	int i, t;
 
+	hds = fbn->hds_thresh < FBNIC_HDR_BYTES_MIN;
+
 	for (i = 0; i < fbn->num_napi; i++) {
 		struct fbnic_napi_vector *nv = fbn->napi[i];
 
 		for (t = 0; t < nv->rxt_count; t++) {
 			struct fbnic_q_triad *qt = &nv->qt[nv->txt_count + t];
 
-			fbnic_config_drop_mode_rcq(nv, &qt->cmpl, tx_pause);
+			fbnic_config_drop_mode_rcq(nv, &qt->cmpl, txp, hds);
 		}
 	}
 }
@@ -2654,20 +2659,18 @@ static void fbnic_enable_rcq(struct fbnic_napi_vector *nv,
 {
 	struct fbnic_net *fbn = netdev_priv(nv->napi.dev);
 	u32 log_size = fls(rcq->size_mask);
-	u32 hds_thresh = fbn->hds_thresh;
 	u32 rcq_ctl = 0;
-
-	fbnic_config_drop_mode_rcq(nv, rcq, fbn->tx_pause);
+	bool hdr_split;
+	u32 hds_thresh;
 
 	/* Force lower bound on MAX_HEADER_BYTES. Below this, all frames should
 	 * be split at L4. It would also result in the frames being split at
 	 * L2/L3 depending on the frame size.
 	 */
-	if (fbn->hds_thresh < FBNIC_HDR_BYTES_MIN) {
-		rcq_ctl = FBNIC_QUEUE_RDE_CTL0_EN_HDR_SPLIT;
-		hds_thresh = FBNIC_HDR_BYTES_MIN;
-	}
+	hdr_split = fbn->hds_thresh < FBNIC_HDR_BYTES_MIN;
+	fbnic_config_drop_mode_rcq(nv, rcq, fbn->tx_pause, hdr_split);
 
+	hds_thresh = max(fbn->hds_thresh, FBNIC_HDR_BYTES_MIN);
 	rcq_ctl |= FIELD_PREP(FBNIC_QUEUE_RDE_CTL1_PADLEN_MASK, FBNIC_RX_PAD) |
 		   FIELD_PREP(FBNIC_QUEUE_RDE_CTL1_MAX_HDR_MASK, hds_thresh) |
 		   FIELD_PREP(FBNIC_QUEUE_RDE_CTL1_PAYLD_OFF_MASK,

-- 
2.47.3

[PATCH net 2/4] eth: fbnic: increase FBNIC_HDR_BYTES_MIN from 128 to 256 bytes

[Bobby Eshleman]

From: Bobby Eshleman <bobbyeshleman@meta.com>

Increase FBNIC_HDR_BYTES_MIN from 128 to 256 bytes. The previous minimum
was too small to guarantee that very long L2+L3+L4 headers always fit
within the header buffer. When EN_HDR_SPLIT is disabled and a packet
exceeds MAX_HEADER_BYTES, splitting occurs at that byte offset instead
of the header boundary, resulting in some of the header landing in the
payload page. The increased minimum ensures headers always fit with the
MAX_HEADER_BYTES cut off and land in the header page.

Fixes: 2b30fc01a6c7 ("eth: fbnic: Add support for HDS configuration")
Signed-off-by: Bobby Eshleman <bobbyeshleman@meta.com>
Acked-by: Mohsin Bashir <mohsin.bashr@gmail.com>
---
 drivers/net/ethernet/meta/fbnic/fbnic_txrx.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/meta/fbnic/fbnic_txrx.h b/drivers/net/ethernet/meta/fbnic/fbnic_txrx.h
index 27776e844e29..51a98f27d5d9 100644
--- a/drivers/net/ethernet/meta/fbnic/fbnic_txrx.h
+++ b/drivers/net/ethernet/meta/fbnic/fbnic_txrx.h
@@ -66,7 +66,7 @@ struct fbnic_net;
 	(4096 - FBNIC_RX_HROOM - FBNIC_RX_TROOM - FBNIC_RX_PAD)
 #define FBNIC_HDS_THRESH_DEFAULT \
 	(1536 - FBNIC_RX_PAD)
-#define FBNIC_HDR_BYTES_MIN		128
+#define FBNIC_HDR_BYTES_MIN		256
 
 struct fbnic_pkt_buff {
 	struct xdp_buff buff;

-- 
2.47.3

[PATCH net 3/4] eth: fbnic: set DMA_HINT_L4 for all flows

[Bobby Eshleman]

From: Bobby Eshleman <bobbyeshleman@meta.com>

fbnic always advertises ETHTOOL_TCP_DATA_SPLIT_ENABLED via ethtool
.get_ringparam. To enable proper splitting for all flow types, even for
IP/Ethernet flows, this patch sets DMA_HINT_L4 unconditionally for all
RSS and NFC flow steering rules. According to the spec, L4 falls back to
L3 if no valid L4 is found, and L3 falls back to L2 if no L3 is found.
This makes sure that the correct header boundary is used regardless of
traffic type. This is important for zero-copy use cases where we must
ensure that all ZC packets are split correctly.

Fixes: 2b30fc01a6c7 ("eth: fbnic: Add support for HDS configuration")
Signed-off-by: Bobby Eshleman <bobbyeshleman@meta.com>
---
 drivers/net/ethernet/meta/fbnic/fbnic_ethtool.c | 3 +++
 drivers/net/ethernet/meta/fbnic/fbnic_rpc.c     | 5 ++---
 2 files changed, 5 insertions(+), 3 deletions(-)

diff --git a/drivers/net/ethernet/meta/fbnic/fbnic_ethtool.c b/drivers/net/ethernet/meta/fbnic/fbnic_ethtool.c
index 693ebdf38705..5edc28ba2955 100644
--- a/drivers/net/ethernet/meta/fbnic/fbnic_ethtool.c
+++ b/drivers/net/ethernet/meta/fbnic/fbnic_ethtool.c
@@ -1142,6 +1142,9 @@ static int fbnic_set_cls_rule_ins(struct fbnic_net *fbn,
 		return -EINVAL;
 	}
 
+	dest |= FIELD_PREP(FBNIC_RPC_ACT_TBL0_DMA_HINT,
+			   FBNIC_RCD_HDR_AL_DMA_HINT_L4);
+
 	/* Write action table values */
 	act_tcam->dest = dest;
 	act_tcam->rss_en_mask = fbnic_flow_hash_2_rss_en_mask(fbn, hash_idx);
diff --git a/drivers/net/ethernet/meta/fbnic/fbnic_rpc.c b/drivers/net/ethernet/meta/fbnic/fbnic_rpc.c
index 7f31e890031c..42a186db43ea 100644
--- a/drivers/net/ethernet/meta/fbnic/fbnic_rpc.c
+++ b/drivers/net/ethernet/meta/fbnic/fbnic_rpc.c
@@ -338,9 +338,8 @@ void fbnic_rss_reinit(struct fbnic_dev *fbd, struct fbnic_net *fbn)
 		else if (tstamp_mask & (1u << flow_type))
 			dest |= FBNIC_RPC_ACT_TBL0_TS_ENA;
 
-		if (act1_value[flow_type] & FBNIC_RPC_TCAM_ACT1_L4_VALID)
-			dest |= FIELD_PREP(FBNIC_RPC_ACT_TBL0_DMA_HINT,
-					   FBNIC_RCD_HDR_AL_DMA_HINT_L4);
+		dest |= FIELD_PREP(FBNIC_RPC_ACT_TBL0_DMA_HINT,
+				   FBNIC_RCD_HDR_AL_DMA_HINT_L4);
 
 		rss_en_mask = fbnic_flow_hash_2_rss_en_mask(fbn, flow_type);
 

-- 
2.47.3

[PATCH net 4/4] selftests: drv-net: add HDS payload sweep test for devmem TCP

[Bobby Eshleman]

From: Bobby Eshleman <bobbyeshleman@meta.com>

Add check_rx_hds test that verifies header/data split works across
payload sizes. The test sweeps payload sizes from 1 byte to 8KB, if any
data propagates up to userspace as SCM_DEVMEM_LINEAR, then the test
fails. This shows that regardless of payload size, ncdevmem's
configuration of hds-thresh to 0 is respected.

Add -L (--fail-on-linear) flag to ncdevmem that causes the receiver to
fail if any SCM_DEVMEM_LINEAR cmsg is received.

Use socat option for fixed block sizing and tcp nodelay to disable
nagle's algo to avoid buffering.

Signed-off-by: Bobby Eshleman <bobbyeshleman@meta.com>
---
 tools/testing/selftests/drivers/net/hw/devmem.py  | 19 ++++++++++++++++++-
 tools/testing/selftests/drivers/net/hw/ncdevmem.c | 11 ++++++++++-
 2 files changed, 28 insertions(+), 2 deletions(-)

diff --git a/tools/testing/selftests/drivers/net/hw/devmem.py b/tools/testing/selftests/drivers/net/hw/devmem.py
index 45c2d49d55b6..ee863e90d1e0 100755
--- a/tools/testing/selftests/drivers/net/hw/devmem.py
+++ b/tools/testing/selftests/drivers/net/hw/devmem.py
@@ -63,12 +63,29 @@ def check_tx_chunks(cfg) -> None:
     ksft_eq(socat.stdout.strip(), "hello\nworld")
 
 
+def check_rx_hds(cfg) -> None:
+    """Test HDS splitting across payload sizes."""
+    require_devmem(cfg)
+
+    for size in [1, 8, 16, 32, 64, 128, 256, 512, 1024, 2048, 4096, 8192]:
+        port = rand_port()
+        listen_cmd = f"{cfg.bin_local} -L -l -f {cfg.ifname} -s {cfg.addr} -p {port}"
+
+        with bkg(listen_cmd, exit_wait=True) as ncdevmem:
+            wait_port_listen(port)
+            cmd(f"dd if=/dev/zero bs={size} count=1 2>/dev/null | " +
+                f"socat -b {size} -u - TCP{cfg.addr_ipver}:{cfg.baddr}:{port},nodelay",
+                host=cfg.remote, shell=True)
+
+        ksft_eq(ncdevmem.ret, 0, f"HDS failed for payload size {size}")
+
+
 def main() -> None:
     with NetDrvEpEnv(__file__) as cfg:
         cfg.bin_local = path.abspath(path.dirname(__file__) + "/ncdevmem")
         cfg.bin_remote = cfg.remote.deploy(cfg.bin_local)
 
-        ksft_run([check_rx, check_tx, check_tx_chunks],
+        ksft_run([check_rx, check_tx, check_tx_chunks, check_rx_hds],
                  args=(cfg, ))
     ksft_exit()
 
diff --git a/tools/testing/selftests/drivers/net/hw/ncdevmem.c b/tools/testing/selftests/drivers/net/hw/ncdevmem.c
index 3288ed04ce08..adc89591c834 100644
--- a/tools/testing/selftests/drivers/net/hw/ncdevmem.c
+++ b/tools/testing/selftests/drivers/net/hw/ncdevmem.c
@@ -97,6 +97,7 @@ static unsigned int ifindex;
 static unsigned int dmabuf_id;
 static uint32_t tx_dmabuf_id;
 static int waittime_ms = 500;
+static bool fail_on_linear;
 
 /* System state loaded by current_config_load() */
 #define MAX_FLOWS	8
@@ -974,6 +975,11 @@ static int do_server(struct memory_buffer *mem)
 					"SCM_DEVMEM_LINEAR. dmabuf_cmsg->frag_size=%u\n",
 					dmabuf_cmsg->frag_size);
 
+				if (fail_on_linear) {
+					pr_err("received SCM_DEVMEM_LINEAR but --fail-on-linear (-L) set");
+					goto err_close_client;
+				}
+
 				continue;
 			}
 
@@ -1397,8 +1403,11 @@ int main(int argc, char *argv[])
 	int is_server = 0, opt;
 	int ret, err = 1;
 
-	while ((opt = getopt(argc, argv, "ls:c:p:v:q:t:f:z:")) != -1) {
+	while ((opt = getopt(argc, argv, "Lls:c:p:v:q:t:f:z:")) != -1) {
 		switch (opt) {
+		case 'L':
+			fail_on_linear = true;
+			break;
 		case 'l':
 			is_server = 1;
 			break;

-- 
2.47.3

Re: [PATCH net 0/4] eth: fbnic: fix and improve header/data split configuration

[patchwork-bot+netdevbpf]

Hello:

This series was applied to netdev/net.git (main)
by Paolo Abeni <pabeni@redhat.com>:

On Wed, 11 Feb 2026 17:00:40 -0800 you wrote:
> This series fixes TCP HDS configuration in the fbnic driver and adds a
> devmem selftest for the intended behavior.
> 
> The issues fixed include setting the correct CSR for EN_HDR_SPLIT,
> adjusting the hds threshold clamp to avoid long headers overflowing into
> the payload page, and configuring the device to use L4/L3/L2 header
> boundaries when present by programming the DMA hint bit unconditionally
> for all steering rule types.
> 
> [...]

Here is the summary with links:
  - [net,1/4] eth: fbnic: set FBNIC_QUEUE_RDE_CTL0_EN_HDR_SPLIT on RDE_CTL0
    https://git.kernel.org/netdev/net/c/bbeb3bfbffe0
  - [net,2/4] eth: fbnic: increase FBNIC_HDR_BYTES_MIN from 128 to 256 bytes
    https://git.kernel.org/netdev/net/c/bd254115f38d
  - [net,3/4] eth: fbnic: set DMA_HINT_L4 for all flows
    https://git.kernel.org/netdev/net/c/0f30a31b55c4
  - [net,4/4] selftests: drv-net: add HDS payload sweep test for devmem TCP
    https://git.kernel.org/netdev/net/c/e7a3c1adc127

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