[PATCH v1 0/5] fix RX data buffer size

[PATCH v1 0/5] fix RX data buffer size

[Wenjun Wu]

This patch set fixes RX data buffer size in ice, i40e, iavf and idpf driver.

1. Limit the maximum buffer size to no more than 16K - 128.
2. Align max buffer size to 128, or replace RTE_ALIGN with
RTE_ALIGN_FLOOR according to [1].

[1] Commit c9c45beb1b97 ("net/iavf: fix Rx queue buffer size alignment")

Wenjun Wu (5):
  net/i40e: fix RX data buffer size
  net/ice: fix RX data buffer size
  net/iavf: fix RX data buffer size
  net/idpf: fix RX data buffer size
  net/cpfl: fix RX data buffer size

 drivers/common/idpf/idpf_common_rxtx.h | 3 +++
 drivers/net/cpfl/cpfl_rxtx.c           | 3 ++-
 drivers/net/i40e/i40e_rxtx.c           | 2 ++
 drivers/net/i40e/i40e_rxtx.h           | 3 +++
 drivers/net/iavf/iavf_rxtx.c           | 1 +
 drivers/net/iavf/iavf_rxtx.h           | 3 +++
 drivers/net/ice/ice_dcf_ethdev.c       | 3 ++-
 drivers/net/ice/ice_rxtx.c             | 3 ++-
 drivers/net/ice/ice_rxtx.h             | 3 +++
 drivers/net/idpf/idpf_rxtx.c           | 6 ++++--
 10 files changed, 25 insertions(+), 5 deletions(-)

-- 
2.34.1

[PATCH v1 1/5] net/i40e: fix RX data buffer size

[Wenjun Wu]

No matter what the mbuf size is, the data buffer size should not
be greater than 16K - 128.

Fixes: 4861cde46116 ("i40e: new poll mode driver")
Cc: stable@dpdk.org

Signed-off-by: Wenjun Wu <wenjun1.wu@intel.com>
---
 drivers/net/i40e/i40e_rxtx.c | 2 ++
 drivers/net/i40e/i40e_rxtx.h | 3 +++
 2 files changed, 5 insertions(+)

diff --git a/drivers/net/i40e/i40e_rxtx.c b/drivers/net/i40e/i40e_rxtx.c
index 788ffb51c2..fbbefb5015 100644
--- a/drivers/net/i40e/i40e_rxtx.c
+++ b/drivers/net/i40e/i40e_rxtx.c
@@ -2904,6 +2904,8 @@ i40e_rx_queue_config(struct i40e_rx_queue *rxq)
 		rxq->rx_hdr_len = 0;
 		rxq->rx_buf_len = RTE_ALIGN_FLOOR(buf_size,
 			(1 << I40E_RXQ_CTX_DBUFF_SHIFT));
+		rxq->rx_buf_len = RTE_MIN(rxq->rx_buf_len,
+					  I40E_RX_MAX_DATA_BUF_SIZE);
 		rxq->hs_mode = i40e_header_split_none;
 		break;
 	}
diff --git a/drivers/net/i40e/i40e_rxtx.h b/drivers/net/i40e/i40e_rxtx.h
index 5e6eecc501..0376c219be 100644
--- a/drivers/net/i40e/i40e_rxtx.h
+++ b/drivers/net/i40e/i40e_rxtx.h
@@ -21,6 +21,9 @@
 /* In none-PXE mode QLEN must be whole number of 32 descriptors. */
 #define	I40E_ALIGN_RING_DESC	32
 
+/* Max data buffer size must be 16K - 128 bytes */
+#define I40E_RX_MAX_DATA_BUF_SIZE	(16 * 1024 - 128)
+
 #define	I40E_MIN_RING_DESC	64
 #define	I40E_MAX_RING_DESC	4096
 
-- 
2.34.1

[PATCH v1 2/5] net/ice: fix RX data buffer size

[Wenjun Wu]

This patch does two fixes.

1. No matter what the mbuf size is, the data buffer size should not
be greater than 16K - 128.

2. Replace RTE_ALIGN with RTE_ALIGN_FLOOR according to [1].

[1] Commit c9c45beb1b97 ("net/iavf: fix Rx queue buffer size alignment")

Fixes: 50370662b727 ("net/ice: support device and queue ops")
Fixes: 1b009275e2c8 ("net/ice: add Rx queue init in DCF")
Cc: stable@dpdk.org

Signed-off-by: Wenjun Wu <wenjun1.wu@intel.com>
---
 drivers/net/ice/ice_dcf_ethdev.c | 3 ++-
 drivers/net/ice/ice_rxtx.c       | 3 ++-
 drivers/net/ice/ice_rxtx.h       | 3 +++
 3 files changed, 7 insertions(+), 2 deletions(-)

diff --git a/drivers/net/ice/ice_dcf_ethdev.c b/drivers/net/ice/ice_dcf_ethdev.c
index dcbf2af5b0..7304ea721c 100644
--- a/drivers/net/ice/ice_dcf_ethdev.c
+++ b/drivers/net/ice/ice_dcf_ethdev.c
@@ -115,7 +115,8 @@ ice_dcf_init_rxq(struct rte_eth_dev *dev, struct ice_rx_queue *rxq)
 
 	buf_size = rte_pktmbuf_data_room_size(rxq->mp) - RTE_PKTMBUF_HEADROOM;
 	rxq->rx_hdr_len = 0;
-	rxq->rx_buf_len = RTE_ALIGN(buf_size, (1 << ICE_RLAN_CTX_DBUF_S));
+	rxq->rx_buf_len = RTE_ALIGN_FLOOR(buf_size, (1 << ICE_RLAN_CTX_DBUF_S));
+	rxq->rx_buf_len = RTE_MIN(rxq->rx_buf_len, ICE_RX_MAX_DATA_BUF_SIZE);
 	max_pkt_len = RTE_MIN(ICE_SUPPORT_CHAIN_NUM * rxq->rx_buf_len,
 			      dev->data->mtu + ICE_ETH_OVERHEAD);
 
diff --git a/drivers/net/ice/ice_rxtx.c b/drivers/net/ice/ice_rxtx.c
index 0ea0045836..560c1a4af7 100644
--- a/drivers/net/ice/ice_rxtx.c
+++ b/drivers/net/ice/ice_rxtx.c
@@ -259,7 +259,8 @@ ice_program_hw_rx_queue(struct ice_rx_queue *rxq)
 	/* Set buffer size as the head split is disabled. */
 	buf_size = (uint16_t)(rte_pktmbuf_data_room_size(rxq->mp) -
 			      RTE_PKTMBUF_HEADROOM);
-	rxq->rx_buf_len = RTE_ALIGN(buf_size, (1 << ICE_RLAN_CTX_DBUF_S));
+	rxq->rx_buf_len = RTE_ALIGN_FLOOR(buf_size, (1 << ICE_RLAN_CTX_DBUF_S));
+	rxq->rx_buf_len = RTE_MIN(rxq->rx_buf_len, ICE_RX_MAX_DATA_BUF_SIZE);
 	rxq->max_pkt_len =
 		RTE_MIN((uint32_t)ICE_SUPPORT_CHAIN_NUM * rxq->rx_buf_len,
 			frame_size);
diff --git a/drivers/net/ice/ice_rxtx.h b/drivers/net/ice/ice_rxtx.h
index 94f6bcf3d1..89569029e1 100644
--- a/drivers/net/ice/ice_rxtx.h
+++ b/drivers/net/ice/ice_rxtx.h
@@ -51,6 +51,9 @@ extern int ice_timestamp_dynfield_offset;
 /* Max header size can be 2K - 64 bytes */
 #define ICE_RX_HDR_BUF_SIZE    (2048 - 64)
 
+/* Max data buffer size must be 16K - 128 bytes */
+#define ICE_RX_MAX_DATA_BUF_SIZE	(16 * 1024 - 128)
+
 #define ICE_HEADER_SPLIT_ENA   BIT(0)
 
 typedef void (*ice_rx_release_mbufs_t)(struct ice_rx_queue *rxq);
-- 
2.34.1

[PATCH v1 3/5] net/iavf: fix RX data buffer size

[Wenjun Wu]

No matter what the mbuf size is, the data buffer size should not
be greater than 16K - 128.

Fixes: 69dd4c3d0898 ("net/avf: enable queue and device")
Cc: stable@dpdk.org

Signed-off-by: Wenjun Wu <wenjun1.wu@intel.com>
---
 drivers/net/iavf/iavf_rxtx.c | 1 +
 drivers/net/iavf/iavf_rxtx.h | 3 +++
 2 files changed, 4 insertions(+)

diff --git a/drivers/net/iavf/iavf_rxtx.c b/drivers/net/iavf/iavf_rxtx.c
index b1d0fbceb6..0db3aabd92 100644
--- a/drivers/net/iavf/iavf_rxtx.c
+++ b/drivers/net/iavf/iavf_rxtx.c
@@ -697,6 +697,7 @@ iavf_dev_rx_queue_setup(struct rte_eth_dev *dev, uint16_t queue_idx,
 
 	len = rte_pktmbuf_data_room_size(rxq->mp) - RTE_PKTMBUF_HEADROOM;
 	rxq->rx_buf_len = RTE_ALIGN_FLOOR(len, (1 << IAVF_RXQ_CTX_DBUFF_SHIFT));
+	rxq->rx_buf_len = RTE_MIN(rxq->rx_buf_len, IAVF_RX_MAX_DATA_BUF_SIZE);
 
 	/* Allocate the software ring. */
 	len = nb_desc + IAVF_RX_MAX_BURST;
diff --git a/drivers/net/iavf/iavf_rxtx.h b/drivers/net/iavf/iavf_rxtx.h
index 09e2127db0..f205a2aaf1 100644
--- a/drivers/net/iavf/iavf_rxtx.h
+++ b/drivers/net/iavf/iavf_rxtx.h
@@ -16,6 +16,9 @@
 /* used for Rx Bulk Allocate */
 #define IAVF_RX_MAX_BURST         32
 
+/* Max data buffer size must be 16K - 128 bytes */
+#define IAVF_RX_MAX_DATA_BUF_SIZE (16 * 1024 - 128)
+
 /* used for Vector PMD */
 #define IAVF_VPMD_RX_MAX_BURST    32
 #define IAVF_VPMD_TX_MAX_BURST    32
-- 
2.34.1

[PATCH v1 4/5] net/idpf: fix RX data buffer size

[Wenjun Wu]

This patch does two fixes.

1. No matter what the mbuf size is, the data buffer size should not
be greater than 16K - 128.

2. Align data buffer size to 128.

Fixes: 9c47c29739a1 ("net/idpf: add Rx queue setup")
Cc: stable@dpdk.org

Signed-off-by: Wenjun Wu <wenjun1.wu@intel.com>
---
 drivers/common/idpf/idpf_common_rxtx.h | 3 +++
 drivers/net/idpf/idpf_rxtx.c           | 6 ++++--
 2 files changed, 7 insertions(+), 2 deletions(-)

diff --git a/drivers/common/idpf/idpf_common_rxtx.h b/drivers/common/idpf/idpf_common_rxtx.h
index 11260d07f9..6cb83fc0a6 100644
--- a/drivers/common/idpf/idpf_common_rxtx.h
+++ b/drivers/common/idpf/idpf_common_rxtx.h
@@ -34,6 +34,9 @@
 #define IDPF_MAX_TSO_FRAME_SIZE	262143
 #define IDPF_TX_MAX_MTU_SEG     10
 
+#define IDPF_RLAN_CTX_DBUF_S	7
+#define IDPF_RX_MAX_DATA_BUF_SIZE	(16 * 1024 - 128)
+
 #define IDPF_TX_CKSUM_OFFLOAD_MASK (		\
 		RTE_MBUF_F_TX_IP_CKSUM |	\
 		RTE_MBUF_F_TX_L4_MASK |		\
diff --git a/drivers/net/idpf/idpf_rxtx.c b/drivers/net/idpf/idpf_rxtx.c
index 414f9a37f6..3e3d81ca6d 100644
--- a/drivers/net/idpf/idpf_rxtx.c
+++ b/drivers/net/idpf/idpf_rxtx.c
@@ -155,7 +155,8 @@ idpf_rx_split_bufq_setup(struct rte_eth_dev *dev, struct idpf_rx_queue *rxq,
 	bufq->adapter = adapter;
 
 	len = rte_pktmbuf_data_room_size(bufq->mp) - RTE_PKTMBUF_HEADROOM;
-	bufq->rx_buf_len = len;
+	bufq->rx_buf_len = RTE_ALIGN_FLOOR(len, (1 << IDPF_RLAN_CTX_DBUF_S));
+	bufq->rx_buf_len = RTE_MIN(bufq->rx_buf_len, IDPF_RX_MAX_DATA_BUF_SIZE);
 
 	/* Allocate a little more to support bulk allocate. */
 	len = nb_desc + IDPF_RX_MAX_BURST;
@@ -275,7 +276,8 @@ idpf_rx_queue_setup(struct rte_eth_dev *dev, uint16_t queue_idx,
 	rxq->offloads = idpf_rx_offload_convert(offloads);
 
 	len = rte_pktmbuf_data_room_size(rxq->mp) - RTE_PKTMBUF_HEADROOM;
-	rxq->rx_buf_len = len;
+	rxq->rx_buf_len = RTE_ALIGN_FLOOR(len, (1 << IDPF_RLAN_CTX_DBUF_S));
+	rxq->rx_buf_len = RTE_MIN(rxq->rx_buf_len, IDPF_RX_MAX_DATA_BUF_SIZE);
 
 	/* Allocate a little more to support bulk allocate. */
 	len = nb_desc + IDPF_RX_MAX_BURST;
-- 
2.34.1

[PATCH v1 5/5] net/cpfl: fix RX data buffer size

[Wenjun Wu]

This patch does two fixes.

1. No matter what the mbuf size is, the data buffer size should not
be greater than 16K - 128.

2. Align data buffer size to 128.

Signed-off-by: Wenjun Wu <wenjun1.wu@intel.com>
---
 drivers/net/cpfl/cpfl_rxtx.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/drivers/net/cpfl/cpfl_rxtx.c b/drivers/net/cpfl/cpfl_rxtx.c
index de59b31b3d..75021c3c54 100644
--- a/drivers/net/cpfl/cpfl_rxtx.c
+++ b/drivers/net/cpfl/cpfl_rxtx.c
@@ -155,7 +155,8 @@ cpfl_rx_split_bufq_setup(struct rte_eth_dev *dev, struct idpf_rx_queue *rxq,
 	bufq->adapter = base;
 
 	len = rte_pktmbuf_data_room_size(bufq->mp) - RTE_PKTMBUF_HEADROOM;
-	bufq->rx_buf_len = len;
+	bufq->rx_buf_len = RTE_ALIGN_FLOOR(len, (1 << IDPF_RLAN_CTX_DBUF_S));
+	bufq->rx_buf_len = RTE_MIN(bufq->rx_buf_len, IDPF_RX_MAX_DATA_BUF_SIZE);
 
 	/* Allocate a little more to support bulk allocate. */
 	len = nb_desc + IDPF_RX_MAX_BURST;
-- 
2.34.1

[PATCH v2 0/5] fix Rx data buffer size

[Wenjun Wu]

This patch set fixes RX data buffer size in ice, i40e, iavf and idpf driver.

1. Limit the maximum buffer size to no more than 16K - 128.
2. Align max buffer size to 128, or replace RTE_ALIGN with
RTE_ALIGN_FLOOR according to [1].

[1] Commit c9c45beb1b97 ("net/iavf: fix Rx queue buffer size alignment")

---
v2: fix commit log

Wenjun Wu (5):
  net/i40e: fix Rx data buffer size
  net/ice: fix Rx data buffer size
  net/iavf: fix Rx data buffer size
  net/idpf: fix Rx data buffer size
  net/cpfl: fix Rx data buffer size

 drivers/common/idpf/idpf_common_rxtx.h | 3 +++
 drivers/net/cpfl/cpfl_rxtx.c           | 3 ++-
 drivers/net/i40e/i40e_rxtx.c           | 2 ++
 drivers/net/i40e/i40e_rxtx.h           | 3 +++
 drivers/net/iavf/iavf_rxtx.c           | 1 +
 drivers/net/iavf/iavf_rxtx.h           | 3 +++
 drivers/net/ice/ice_dcf_ethdev.c       | 3 ++-
 drivers/net/ice/ice_rxtx.c             | 3 ++-
 drivers/net/ice/ice_rxtx.h             | 3 +++
 drivers/net/idpf/idpf_rxtx.c           | 6 ++++--
 10 files changed, 25 insertions(+), 5 deletions(-)

-- 
2.34.1

[PATCH v2 1/5] net/i40e: fix Rx data buffer size

[Wenjun Wu]

No matter what the mbuf size is, the data buffer size should not
be greater than 16K - 128.

Fixes: 4861cde46116 ("i40e: new poll mode driver")
Cc: stable@dpdk.org

Signed-off-by: Wenjun Wu <wenjun1.wu@intel.com>
---
 drivers/net/i40e/i40e_rxtx.c | 2 ++
 drivers/net/i40e/i40e_rxtx.h | 3 +++
 2 files changed, 5 insertions(+)

diff --git a/drivers/net/i40e/i40e_rxtx.c b/drivers/net/i40e/i40e_rxtx.c
index 788ffb51c2..fbbefb5015 100644
--- a/drivers/net/i40e/i40e_rxtx.c
+++ b/drivers/net/i40e/i40e_rxtx.c
@@ -2904,6 +2904,8 @@ i40e_rx_queue_config(struct i40e_rx_queue *rxq)
 		rxq->rx_hdr_len = 0;
 		rxq->rx_buf_len = RTE_ALIGN_FLOOR(buf_size,
 			(1 << I40E_RXQ_CTX_DBUFF_SHIFT));
+		rxq->rx_buf_len = RTE_MIN(rxq->rx_buf_len,
+					  I40E_RX_MAX_DATA_BUF_SIZE);
 		rxq->hs_mode = i40e_header_split_none;
 		break;
 	}
diff --git a/drivers/net/i40e/i40e_rxtx.h b/drivers/net/i40e/i40e_rxtx.h
index 5e6eecc501..0376c219be 100644
--- a/drivers/net/i40e/i40e_rxtx.h
+++ b/drivers/net/i40e/i40e_rxtx.h
@@ -21,6 +21,9 @@
 /* In none-PXE mode QLEN must be whole number of 32 descriptors. */
 #define	I40E_ALIGN_RING_DESC	32
 
+/* Max data buffer size must be 16K - 128 bytes */
+#define I40E_RX_MAX_DATA_BUF_SIZE	(16 * 1024 - 128)
+
 #define	I40E_MIN_RING_DESC	64
 #define	I40E_MAX_RING_DESC	4096
 
-- 
2.34.1

[PATCH v2 2/5] net/ice: fix Rx data buffer size

[Wenjun Wu]

This patch does two fixes.

1. No matter what the mbuf size is, the data buffer size should not
be greater than 16K - 128.

2. Replace RTE_ALIGN with RTE_ALIGN_FLOOR according to [1].

[1] Commit c9c45beb1b97 ("net/iavf: fix Rx queue buffer size alignment")

Fixes: 50370662b727 ("net/ice: support device and queue ops")
Fixes: 1b009275e2c8 ("net/ice: add Rx queue init in DCF")
Cc: stable@dpdk.org

Signed-off-by: Wenjun Wu <wenjun1.wu@intel.com>
---
 drivers/net/ice/ice_dcf_ethdev.c | 3 ++-
 drivers/net/ice/ice_rxtx.c       | 3 ++-
 drivers/net/ice/ice_rxtx.h       | 3 +++
 3 files changed, 7 insertions(+), 2 deletions(-)

diff --git a/drivers/net/ice/ice_dcf_ethdev.c b/drivers/net/ice/ice_dcf_ethdev.c
index dcbf2af5b0..7304ea721c 100644
--- a/drivers/net/ice/ice_dcf_ethdev.c
+++ b/drivers/net/ice/ice_dcf_ethdev.c
@@ -115,7 +115,8 @@ ice_dcf_init_rxq(struct rte_eth_dev *dev, struct ice_rx_queue *rxq)
 
 	buf_size = rte_pktmbuf_data_room_size(rxq->mp) - RTE_PKTMBUF_HEADROOM;
 	rxq->rx_hdr_len = 0;
-	rxq->rx_buf_len = RTE_ALIGN(buf_size, (1 << ICE_RLAN_CTX_DBUF_S));
+	rxq->rx_buf_len = RTE_ALIGN_FLOOR(buf_size, (1 << ICE_RLAN_CTX_DBUF_S));
+	rxq->rx_buf_len = RTE_MIN(rxq->rx_buf_len, ICE_RX_MAX_DATA_BUF_SIZE);
 	max_pkt_len = RTE_MIN(ICE_SUPPORT_CHAIN_NUM * rxq->rx_buf_len,
 			      dev->data->mtu + ICE_ETH_OVERHEAD);
 
diff --git a/drivers/net/ice/ice_rxtx.c b/drivers/net/ice/ice_rxtx.c
index 0ea0045836..560c1a4af7 100644
--- a/drivers/net/ice/ice_rxtx.c
+++ b/drivers/net/ice/ice_rxtx.c
@@ -259,7 +259,8 @@ ice_program_hw_rx_queue(struct ice_rx_queue *rxq)
 	/* Set buffer size as the head split is disabled. */
 	buf_size = (uint16_t)(rte_pktmbuf_data_room_size(rxq->mp) -
 			      RTE_PKTMBUF_HEADROOM);
-	rxq->rx_buf_len = RTE_ALIGN(buf_size, (1 << ICE_RLAN_CTX_DBUF_S));
+	rxq->rx_buf_len = RTE_ALIGN_FLOOR(buf_size, (1 << ICE_RLAN_CTX_DBUF_S));
+	rxq->rx_buf_len = RTE_MIN(rxq->rx_buf_len, ICE_RX_MAX_DATA_BUF_SIZE);
 	rxq->max_pkt_len =
 		RTE_MIN((uint32_t)ICE_SUPPORT_CHAIN_NUM * rxq->rx_buf_len,
 			frame_size);
diff --git a/drivers/net/ice/ice_rxtx.h b/drivers/net/ice/ice_rxtx.h
index 94f6bcf3d1..89569029e1 100644
--- a/drivers/net/ice/ice_rxtx.h
+++ b/drivers/net/ice/ice_rxtx.h
@@ -51,6 +51,9 @@ extern int ice_timestamp_dynfield_offset;
 /* Max header size can be 2K - 64 bytes */
 #define ICE_RX_HDR_BUF_SIZE    (2048 - 64)
 
+/* Max data buffer size must be 16K - 128 bytes */
+#define ICE_RX_MAX_DATA_BUF_SIZE	(16 * 1024 - 128)
+
 #define ICE_HEADER_SPLIT_ENA   BIT(0)
 
 typedef void (*ice_rx_release_mbufs_t)(struct ice_rx_queue *rxq);
-- 
2.34.1

[PATCH v2 3/5] net/iavf: fix Rx data buffer size

[Wenjun Wu]

No matter what the mbuf size is, the data buffer size should not
be greater than 16K - 128.

Fixes: 69dd4c3d0898 ("net/avf: enable queue and device")
Cc: stable@dpdk.org

Signed-off-by: Wenjun Wu <wenjun1.wu@intel.com>
---
 drivers/net/iavf/iavf_rxtx.c | 1 +
 drivers/net/iavf/iavf_rxtx.h | 3 +++
 2 files changed, 4 insertions(+)

diff --git a/drivers/net/iavf/iavf_rxtx.c b/drivers/net/iavf/iavf_rxtx.c
index b1d0fbceb6..0db3aabd92 100644
--- a/drivers/net/iavf/iavf_rxtx.c
+++ b/drivers/net/iavf/iavf_rxtx.c
@@ -697,6 +697,7 @@ iavf_dev_rx_queue_setup(struct rte_eth_dev *dev, uint16_t queue_idx,
 
 	len = rte_pktmbuf_data_room_size(rxq->mp) - RTE_PKTMBUF_HEADROOM;
 	rxq->rx_buf_len = RTE_ALIGN_FLOOR(len, (1 << IAVF_RXQ_CTX_DBUFF_SHIFT));
+	rxq->rx_buf_len = RTE_MIN(rxq->rx_buf_len, IAVF_RX_MAX_DATA_BUF_SIZE);
 
 	/* Allocate the software ring. */
 	len = nb_desc + IAVF_RX_MAX_BURST;
diff --git a/drivers/net/iavf/iavf_rxtx.h b/drivers/net/iavf/iavf_rxtx.h
index 09e2127db0..f205a2aaf1 100644
--- a/drivers/net/iavf/iavf_rxtx.h
+++ b/drivers/net/iavf/iavf_rxtx.h
@@ -16,6 +16,9 @@
 /* used for Rx Bulk Allocate */
 #define IAVF_RX_MAX_BURST         32
 
+/* Max data buffer size must be 16K - 128 bytes */
+#define IAVF_RX_MAX_DATA_BUF_SIZE (16 * 1024 - 128)
+
 /* used for Vector PMD */
 #define IAVF_VPMD_RX_MAX_BURST    32
 #define IAVF_VPMD_TX_MAX_BURST    32
-- 
2.34.1

[PATCH v2 4/5] net/idpf: fix Rx data buffer size

[Wenjun Wu]

This patch does two fixes.

1. No matter what the mbuf size is, the data buffer size should not
be greater than 16K - 128.

2. Align data buffer size to 128.

Fixes: 9c47c29739a1 ("net/idpf: add Rx queue setup")
Cc: stable@dpdk.org

Signed-off-by: Wenjun Wu <wenjun1.wu@intel.com>
---
 drivers/common/idpf/idpf_common_rxtx.h | 3 +++
 drivers/net/idpf/idpf_rxtx.c           | 6 ++++--
 2 files changed, 7 insertions(+), 2 deletions(-)

diff --git a/drivers/common/idpf/idpf_common_rxtx.h b/drivers/common/idpf/idpf_common_rxtx.h
index 11260d07f9..6cb83fc0a6 100644
--- a/drivers/common/idpf/idpf_common_rxtx.h
+++ b/drivers/common/idpf/idpf_common_rxtx.h
@@ -34,6 +34,9 @@
 #define IDPF_MAX_TSO_FRAME_SIZE	262143
 #define IDPF_TX_MAX_MTU_SEG     10
 
+#define IDPF_RLAN_CTX_DBUF_S	7
+#define IDPF_RX_MAX_DATA_BUF_SIZE	(16 * 1024 - 128)
+
 #define IDPF_TX_CKSUM_OFFLOAD_MASK (		\
 		RTE_MBUF_F_TX_IP_CKSUM |	\
 		RTE_MBUF_F_TX_L4_MASK |		\
diff --git a/drivers/net/idpf/idpf_rxtx.c b/drivers/net/idpf/idpf_rxtx.c
index 414f9a37f6..3e3d81ca6d 100644
--- a/drivers/net/idpf/idpf_rxtx.c
+++ b/drivers/net/idpf/idpf_rxtx.c
@@ -155,7 +155,8 @@ idpf_rx_split_bufq_setup(struct rte_eth_dev *dev, struct idpf_rx_queue *rxq,
 	bufq->adapter = adapter;
 
 	len = rte_pktmbuf_data_room_size(bufq->mp) - RTE_PKTMBUF_HEADROOM;
-	bufq->rx_buf_len = len;
+	bufq->rx_buf_len = RTE_ALIGN_FLOOR(len, (1 << IDPF_RLAN_CTX_DBUF_S));
+	bufq->rx_buf_len = RTE_MIN(bufq->rx_buf_len, IDPF_RX_MAX_DATA_BUF_SIZE);
 
 	/* Allocate a little more to support bulk allocate. */
 	len = nb_desc + IDPF_RX_MAX_BURST;
@@ -275,7 +276,8 @@ idpf_rx_queue_setup(struct rte_eth_dev *dev, uint16_t queue_idx,
 	rxq->offloads = idpf_rx_offload_convert(offloads);
 
 	len = rte_pktmbuf_data_room_size(rxq->mp) - RTE_PKTMBUF_HEADROOM;
-	rxq->rx_buf_len = len;
+	rxq->rx_buf_len = RTE_ALIGN_FLOOR(len, (1 << IDPF_RLAN_CTX_DBUF_S));
+	rxq->rx_buf_len = RTE_MIN(rxq->rx_buf_len, IDPF_RX_MAX_DATA_BUF_SIZE);
 
 	/* Allocate a little more to support bulk allocate. */
 	len = nb_desc + IDPF_RX_MAX_BURST;
-- 
2.34.1

[PATCH v2 5/5] net/cpfl: fix Rx data buffer size

[Wenjun Wu]

This patch does two fixes.

1. No matter what the mbuf size is, the data buffer size should not
be greater than 16K - 128.

2. Align data buffer size to 128.

Fixes: 119834846e93 ("net/cpfl: support Rx queue setup")
Cc: stable@dpdk.org

Signed-off-by: Wenjun Wu <wenjun1.wu@intel.com>
---
 drivers/net/cpfl/cpfl_rxtx.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/drivers/net/cpfl/cpfl_rxtx.c b/drivers/net/cpfl/cpfl_rxtx.c
index de59b31b3d..75021c3c54 100644
--- a/drivers/net/cpfl/cpfl_rxtx.c
+++ b/drivers/net/cpfl/cpfl_rxtx.c
@@ -155,7 +155,8 @@ cpfl_rx_split_bufq_setup(struct rte_eth_dev *dev, struct idpf_rx_queue *rxq,
 	bufq->adapter = base;
 
 	len = rte_pktmbuf_data_room_size(bufq->mp) - RTE_PKTMBUF_HEADROOM;
-	bufq->rx_buf_len = len;
+	bufq->rx_buf_len = RTE_ALIGN_FLOOR(len, (1 << IDPF_RLAN_CTX_DBUF_S));
+	bufq->rx_buf_len = RTE_MIN(bufq->rx_buf_len, IDPF_RX_MAX_DATA_BUF_SIZE);
 
 	/* Allocate a little more to support bulk allocate. */
 	len = nb_desc + IDPF_RX_MAX_BURST;
-- 
2.34.1

RE: [PATCH v2 0/5] fix Rx data buffer size

[Zhang, Qi Z]

> -----Original Message-----
> From: Wu, Wenjun1 <wenjun1.wu@intel.com>
> Sent: Friday, April 14, 2023 1:48 PM
> To: dev@dpdk.org; Zhang, Yuying <yuying.zhang@intel.com>; Xing, Beilei
> <beilei.xing@intel.com>; Wu, Jingjing <jingjing.wu@intel.com>; Yang,
> Qiming <qiming.yang@intel.com>; Zhang, Qi Z <qi.z.zhang@intel.com>
> Cc: Wu, Wenjun1 <wenjun1.wu@intel.com>
> Subject: [PATCH v2 0/5] fix Rx data buffer size
> 
> This patch set fixes RX data buffer size in ice, i40e, iavf and idpf driver.
> 
> 1. Limit the maximum buffer size to no more than 16K - 128.

Confirmed with author, this target to align with hardware spec, added related explanation in commit log.

> 2. Align max buffer size to 128, or replace RTE_ALIGN with
> RTE_ALIGN_FLOOR according to [1].
> 
> [1] Commit c9c45beb1b97 ("net/iavf: fix Rx queue buffer size alignment")
> 
> ---
> v2: fix commit log
> 
> Wenjun Wu (5):
>   net/i40e: fix Rx data buffer size
>   net/ice: fix Rx data buffer size
>   net/iavf: fix Rx data buffer size
>   net/idpf: fix Rx data buffer size
>   net/cpfl: fix Rx data buffer size
> 
>  drivers/common/idpf/idpf_common_rxtx.h | 3 +++
>  drivers/net/cpfl/cpfl_rxtx.c           | 3 ++-
>  drivers/net/i40e/i40e_rxtx.c           | 2 ++
>  drivers/net/i40e/i40e_rxtx.h           | 3 +++
>  drivers/net/iavf/iavf_rxtx.c           | 1 +
>  drivers/net/iavf/iavf_rxtx.h           | 3 +++
>  drivers/net/ice/ice_dcf_ethdev.c       | 3 ++-
>  drivers/net/ice/ice_rxtx.c             | 3 ++-
>  drivers/net/ice/ice_rxtx.h             | 3 +++
>  drivers/net/idpf/idpf_rxtx.c           | 6 ++++--
>  10 files changed, 25 insertions(+), 5 deletions(-)
> 
> --
> 2.34.1

Acked-by: Qi Zhang <qi.z.zhang@intel.com>

Applied to dpdk-next-net-intel.

Thanks
Qi