[dpdk-dev] [PATCH] net/ice: fix TSO pkt exceeds allowed buf size issue

[dpdk-dev] [PATCH] net/ice: fix TSO pkt exceeds allowed buf size issue

[Xiaoyun Li]

Hardware limits that max buffer size per tx descriptor should be
(16K-1)B. So when TSO enabled, the mbuf data size may exceed the
limit and cause malicious behavior to the NIC. This patch fixes
this issue by using more tx descs for this kind of large buffer.

Fixes: 17c7d0f9d6a4 ("net/ice: support basic Rx/Tx")
Cc: stable@dpdk.org

Signed-off-by: Xiaoyun Li <xiaoyun.li@intel.com>
---
 drivers/net/ice/ice_rxtx.c | 59 ++++++++++++++++++++++++++++++++++----
 1 file changed, 54 insertions(+), 5 deletions(-)

diff --git a/drivers/net/ice/ice_rxtx.c b/drivers/net/ice/ice_rxtx.c
index b4f5367c5..c3d549ef4 100644
--- a/drivers/net/ice/ice_rxtx.c
+++ b/drivers/net/ice/ice_rxtx.c
@@ -2421,6 +2421,24 @@ ice_set_tso_ctx(struct rte_mbuf *mbuf, union ice_tx_offload tx_offload)
 	return ctx_desc;
 }
 
+/* HW requires that TX buffer size ranges from 1B up to (16K-1)B. */
+#define ICE_MAX_DATA_PER_TXD \
+	(ICE_TXD_QW1_TX_BUF_SZ_M >> ICE_TXD_QW1_TX_BUF_SZ_S)
+/* Calculate the number of TX descriptors needed for each pkt */
+static inline uint16_t
+ice_calc_pkt_desc(struct rte_mbuf *tx_pkt)
+{
+	struct rte_mbuf *txd = tx_pkt;
+	uint16_t count = 0;
+
+	while (txd != NULL) {
+		count += DIV_ROUND_UP(txd->data_len, ICE_MAX_DATA_PER_TXD);
+		txd = txd->next;
+	}
+
+	return count;
+}
+
 uint16_t
 ice_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts, uint16_t nb_pkts)
 {
@@ -2440,6 +2458,7 @@ ice_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts, uint16_t nb_pkts)
 	uint32_t td_offset = 0;
 	uint32_t td_tag = 0;
 	uint16_t tx_last;
+	uint16_t slen;
 	uint64_t buf_dma_addr;
 	uint64_t ol_flags;
 	union ice_tx_offload tx_offload = {0};
@@ -2471,8 +2490,15 @@ ice_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts, uint16_t nb_pkts)
 		/* The number of descriptors that must be allocated for
 		 * a packet equals to the number of the segments of that
 		 * packet plus the number of context descriptor if needed.
+		 * Recalculate the needed tx descs when TSO enabled in case
+		 * the mbuf data size exceeds max data size that hw allows
+		 * per tx desc.
 		 */
-		nb_used = (uint16_t)(tx_pkt->nb_segs + nb_ctx);
+		if (ol_flags & PKT_TX_TCP_SEG)
+			nb_used = (uint16_t)(ice_calc_pkt_desc(tx_pkt) +
+					     nb_ctx);
+		else
+			nb_used = (uint16_t)(tx_pkt->nb_segs + nb_ctx);
 		tx_last = (uint16_t)(tx_id + nb_used - 1);
 
 		/* Circular ring */
@@ -2562,15 +2588,38 @@ ice_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts, uint16_t nb_pkts)
 			txe->mbuf = m_seg;
 
 			/* Setup TX Descriptor */
+			slen = m_seg->data_len;
 			buf_dma_addr = rte_mbuf_data_iova(m_seg);
+
+			while ((ol_flags & PKT_TX_TCP_SEG) &&
+				unlikely(slen > ICE_MAX_DATA_PER_TXD)) {
+				txd->buffer_addr =
+					rte_cpu_to_le_64(buf_dma_addr);
+				txd->cmd_type_offset_bsz =
+				rte_cpu_to_le_64(ICE_TX_DESC_DTYPE_DATA |
+				((uint64_t)td_cmd << ICE_TXD_QW1_CMD_S) |
+				((uint64_t)td_offset << ICE_TXD_QW1_OFFSET_S) |
+				((uint64_t)ICE_MAX_DATA_PER_TXD <<
+				 ICE_TXD_QW1_TX_BUF_SZ_S) |
+				((uint64_t)td_tag << ICE_TXD_QW1_L2TAG1_S));
+
+				buf_dma_addr += ICE_MAX_DATA_PER_TXD;
+				slen -= ICE_MAX_DATA_PER_TXD;
+
+				txe->last_id = tx_last;
+				tx_id = txe->next_id;
+				txe = txn;
+				txd = &txr[tx_id];
+				txn = &sw_ring[txe->next_id];
+			}
+
 			txd->buf_addr = rte_cpu_to_le_64(buf_dma_addr);
 			txd->cmd_type_offset_bsz =
 				rte_cpu_to_le_64(ICE_TX_DESC_DTYPE_DATA |
-				((uint64_t)td_cmd  << ICE_TXD_QW1_CMD_S) |
+				((uint64_t)td_cmd << ICE_TXD_QW1_CMD_S) |
 				((uint64_t)td_offset << ICE_TXD_QW1_OFFSET_S) |
-				((uint64_t)m_seg->data_len  <<
-				 ICE_TXD_QW1_TX_BUF_SZ_S) |
-				((uint64_t)td_tag  << ICE_TXD_QW1_L2TAG1_S));
+				((uint64_t)slen << ICE_TXD_QW1_TX_BUF_SZ_S) |
+				((uint64_t)td_tag << ICE_TXD_QW1_L2TAG1_S));
 
 			txe->last_id = tx_last;
 			tx_id = txe->next_id;
-- 
2.17.1

[dpdk-dev] [PATCH v2] net/ice: fix TSO pkt exceeds allowed buf size issue

[Xiaoyun Li]

Hardware limits that max buffer size per tx descriptor should be
(16K-1)B. So when TSO enabled, the mbuf data size may exceed the
limit and cause malicious behavior to the NIC. This patch fixes
this issue by using more tx descs for this kind of large buffer.

Fixes: 17c7d0f9d6a4 ("net/ice: support basic Rx/Tx")
Cc: stable@dpdk.org

Signed-off-by: Xiaoyun Li <xiaoyun.li@intel.com>
---
v2:
 * Fixed several typos.
---
 drivers/net/ice/ice_rxtx.c | 58 ++++++++++++++++++++++++++++++++++----
 1 file changed, 53 insertions(+), 5 deletions(-)

diff --git a/drivers/net/ice/ice_rxtx.c b/drivers/net/ice/ice_rxtx.c
index b4f5367c5..ccc2c9339 100644
--- a/drivers/net/ice/ice_rxtx.c
+++ b/drivers/net/ice/ice_rxtx.c
@@ -2421,6 +2421,24 @@ ice_set_tso_ctx(struct rte_mbuf *mbuf, union ice_tx_offload tx_offload)
 	return ctx_desc;
 }
 
+/* HW requires that TX buffer size ranges from 1B up to (16K-1)B. */
+#define ICE_MAX_DATA_PER_TXD \
+	(ICE_TXD_QW1_TX_BUF_SZ_M >> ICE_TXD_QW1_TX_BUF_SZ_S)
+/* Calculate the number of TX descriptors needed for each pkt */
+static inline uint16_t
+ice_calc_pkt_desc(struct rte_mbuf *tx_pkt)
+{
+	struct rte_mbuf *txd = tx_pkt;
+	uint16_t count = 0;
+
+	while (txd != NULL) {
+		count += DIV_ROUND_UP(txd->data_len, ICE_MAX_DATA_PER_TXD);
+		txd = txd->next;
+	}
+
+	return count;
+}
+
 uint16_t
 ice_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts, uint16_t nb_pkts)
 {
@@ -2440,6 +2458,7 @@ ice_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts, uint16_t nb_pkts)
 	uint32_t td_offset = 0;
 	uint32_t td_tag = 0;
 	uint16_t tx_last;
+	uint16_t slen;
 	uint64_t buf_dma_addr;
 	uint64_t ol_flags;
 	union ice_tx_offload tx_offload = {0};
@@ -2471,8 +2490,15 @@ ice_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts, uint16_t nb_pkts)
 		/* The number of descriptors that must be allocated for
 		 * a packet equals to the number of the segments of that
 		 * packet plus the number of context descriptor if needed.
+		 * Recalculate the needed tx descs when TSO enabled in case
+		 * the mbuf data size exceeds max data size that hw allows
+		 * per tx desc.
 		 */
-		nb_used = (uint16_t)(tx_pkt->nb_segs + nb_ctx);
+		if (ol_flags & PKT_TX_TCP_SEG)
+			nb_used = (uint16_t)(ice_calc_pkt_desc(tx_pkt) +
+					     nb_ctx);
+		else
+			nb_used = (uint16_t)(tx_pkt->nb_segs + nb_ctx);
 		tx_last = (uint16_t)(tx_id + nb_used - 1);
 
 		/* Circular ring */
@@ -2562,15 +2588,37 @@ ice_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts, uint16_t nb_pkts)
 			txe->mbuf = m_seg;
 
 			/* Setup TX Descriptor */
+			slen = m_seg->data_len;
 			buf_dma_addr = rte_mbuf_data_iova(m_seg);
+
+			while ((ol_flags & PKT_TX_TCP_SEG) &&
+				unlikely(slen > ICE_MAX_DATA_PER_TXD)) {
+				txd->buf_addr = rte_cpu_to_le_64(buf_dma_addr);
+				txd->cmd_type_offset_bsz =
+				rte_cpu_to_le_64(ICE_TX_DESC_DTYPE_DATA |
+				((uint64_t)td_cmd << ICE_TXD_QW1_CMD_S) |
+				((uint64_t)td_offset << ICE_TXD_QW1_OFFSET_S) |
+				((uint64_t)ICE_MAX_DATA_PER_TXD <<
+				 ICE_TXD_QW1_TX_BUF_SZ_S) |
+				((uint64_t)td_tag << ICE_TXD_QW1_L2TAG1_S));
+
+				buf_dma_addr += ICE_MAX_DATA_PER_TXD;
+				slen -= ICE_MAX_DATA_PER_TXD;
+
+				txe->last_id = tx_last;
+				tx_id = txe->next_id;
+				txe = txn;
+				txd = &tx_ring[tx_id];
+				txn = &sw_ring[txe->next_id];
+			}
+
 			txd->buf_addr = rte_cpu_to_le_64(buf_dma_addr);
 			txd->cmd_type_offset_bsz =
 				rte_cpu_to_le_64(ICE_TX_DESC_DTYPE_DATA |
-				((uint64_t)td_cmd  << ICE_TXD_QW1_CMD_S) |
+				((uint64_t)td_cmd << ICE_TXD_QW1_CMD_S) |
 				((uint64_t)td_offset << ICE_TXD_QW1_OFFSET_S) |
-				((uint64_t)m_seg->data_len  <<
-				 ICE_TXD_QW1_TX_BUF_SZ_S) |
-				((uint64_t)td_tag  << ICE_TXD_QW1_L2TAG1_S));
+				((uint64_t)slen << ICE_TXD_QW1_TX_BUF_SZ_S) |
+				((uint64_t)td_tag << ICE_TXD_QW1_L2TAG1_S));
 
 			txe->last_id = tx_last;
 			tx_id = txe->next_id;
-- 
2.17.1

Re: [dpdk-dev] [PATCH v2] net/ice: fix TSO pkt exceeds allowed buf size issue

[Zhang, Qi Z]

> -----Original Message-----
> From: Li, Xiaoyun <xiaoyun.li@intel.com>
> Sent: Thursday, December 26, 2019 2:54 PM
> To: Zhang, Qi Z <qi.z.zhang@intel.com>; Xing, Beilei <beilei.xing@intel.com>;
> Yang, Qiming <qiming.yang@intel.com>; Lu, Wenzhuo
> <wenzhuo.lu@intel.com>; Loftus, Ciara <ciara.loftus@intel.com>;
> dev@dpdk.org
> Cc: Li, Xiaoyun <xiaoyun.li@intel.com>; stable@dpdk.org
> Subject: [PATCH v2] net/ice: fix TSO pkt exceeds allowed buf size issue
> 
> Hardware limits that max buffer size per tx descriptor should be (16K-1)B. So
> when TSO enabled, the mbuf data size may exceed the limit and cause
> malicious behavior to the NIC. This patch fixes this issue by using more tx descs
> for this kind of large buffer.
> 
> Fixes: 17c7d0f9d6a4 ("net/ice: support basic Rx/Tx")
> Cc: stable@dpdk.org
> 
> Signed-off-by: Xiaoyun Li <xiaoyun.li@intel.com>

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

Re: [dpdk-dev] [PATCH v2] net/ice: fix TSO pkt exceeds allowed buf size issue

[Ye Xiaolong]

On 12/26, Xiaoyun Li wrote:
>Hardware limits that max buffer size per tx descriptor should be
>(16K-1)B. So when TSO enabled, the mbuf data size may exceed the
>limit and cause malicious behavior to the NIC. This patch fixes
>this issue by using more tx descs for this kind of large buffer.
>
>Fixes: 17c7d0f9d6a4 ("net/ice: support basic Rx/Tx")
>Cc: stable@dpdk.org
>
>Signed-off-by: Xiaoyun Li <xiaoyun.li@intel.com>
>---

Applied to dpdk-next-net-intel, Thanks.

Re: [dpdk-dev] [PATCH] net/ice: fix TSO pkt exceeds allowed buf size issue

[Zhang, Qi Z]

> -----Original Message-----
> From: Li, Xiaoyun <xiaoyun.li@intel.com>
> Sent: Thursday, December 26, 2019 2:43 PM
> To: Zhang, Qi Z <qi.z.zhang@intel.com>; Xing, Beilei <beilei.xing@intel.com>;
> Yang, Qiming <qiming.yang@intel.com>; Lu, Wenzhuo
> <wenzhuo.lu@intel.com>; Loftus, Ciara <ciara.loftus@intel.com>;
> dev@dpdk.org
> Cc: Li, Xiaoyun <xiaoyun.li@intel.com>; stable@dpdk.org
> Subject: [PATCH] net/ice: fix TSO pkt exceeds allowed buf size issue
> 
> Hardware limits that max buffer size per tx descriptor should be (16K-1)B. So
> when TSO enabled, the mbuf data size may exceed the limit and cause
> malicious behavior to the NIC. This patch fixes this issue by using more tx descs
> for this kind of large buffer.
> 
> Fixes: 17c7d0f9d6a4 ("net/ice: support basic Rx/Tx")
> Cc: stable@dpdk.org
> 
> Signed-off-by: Xiaoyun Li <xiaoyun.li@intel.com>

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