4 files changed, 61 insertions, 29 deletions

diff --git a/drivers/scsi/cxgbi/cxgb3i/cxgb3i.c b/drivers/scsi/cxgbi/cxgb3i/cxgb3i.c
index 1880eb6c68f7..7b09e7ddf35e 100644
--- a/drivers/scsi/cxgbi/cxgb3i/cxgb3i.c
+++ b/drivers/scsi/cxgbi/cxgb3i/cxgb3i.c
@@ -354,7 +354,7 @@ static inline void make_tx_data_wr(struct cxgbi_sock *csk, struct sk_buff *skb,
 	struct l2t_entry *l2t = csk->l2t;
 
 	skb_reset_transport_header(skb);
-	req = (struct tx_data_wr *)__skb_push(skb, sizeof(*req));
+	req = __skb_push(skb, sizeof(*req));
 	req->wr_hi = htonl(V_WR_OP(FW_WROPCODE_OFLD_TX_DATA) |
 			(req_completion ? F_WR_COMPL : 0));
 	req->wr_lo = htonl(V_WR_TID(csk->tid));
diff --git a/drivers/scsi/cxgbi/cxgb4i/cxgb4i.c b/drivers/scsi/cxgbi/cxgb4i/cxgb4i.c
index 1076c1578322..a69a9ac836f5 100644
--- a/drivers/scsi/cxgbi/cxgb4i/cxgb4i.c
+++ b/drivers/scsi/cxgbi/cxgb4i/cxgb4i.c
@@ -644,7 +644,7 @@ static inline void make_tx_data_wr(struct cxgbi_sock *csk, struct sk_buff *skb,
 	unsigned int wr_ulp_mode = 0, val;
 	bool imm = is_ofld_imm(skb);
 
-	req = (struct fw_ofld_tx_data_wr *)__skb_push(skb, sizeof(*req));
+	req = __skb_push(skb, sizeof(*req));
 
 	if (imm) {
 		req->op_to_immdlen = htonl(FW_WR_OP_V(FW_OFLD_TX_DATA_WR) |
@@ -806,7 +806,7 @@ static void do_act_establish(struct cxgbi_device *cdev, struct sk_buff *skb)
 
 	cxgbi_sock_get(csk);
 	csk->tid = tid;
-	cxgb4_insert_tid(lldi->tids, csk, tid);
+	cxgb4_insert_tid(lldi->tids, csk, tid, csk->csk_family);
 	cxgbi_sock_set_flag(csk, CTPF_HAS_TID);
 
 	free_atid(csk);
@@ -956,7 +956,8 @@ static void do_act_open_rpl(struct cxgbi_device *cdev, struct sk_buff *skb)
 	if (status && status != CPL_ERR_TCAM_FULL &&
 	    status != CPL_ERR_CONN_EXIST &&
 	    status != CPL_ERR_ARP_MISS)
-		cxgb4_remove_tid(lldi->tids, csk->port_id, GET_TID(rpl));
+		cxgb4_remove_tid(lldi->tids, csk->port_id, GET_TID(rpl),
+				 csk->csk_family);
 
 	cxgbi_sock_get(csk);
 	spin_lock_bh(&csk->lock);
@@ -1590,12 +1591,12 @@ static void release_offload_resources(struct cxgbi_sock *csk)
 		free_atid(csk);
 	else if (cxgbi_sock_flag(csk, CTPF_HAS_TID)) {
 		lldi = cxgbi_cdev_priv(csk->cdev);
-		cxgb4_remove_tid(lldi->tids, 0, csk->tid);
+		cxgb4_remove_tid(lldi->tids, 0, csk->tid,
+				 csk->csk_family);
 		cxgbi_sock_clear_flag(csk, CTPF_HAS_TID);
 		cxgbi_sock_put(csk);
 	}
 	csk->dst = NULL;
-	csk->cdev = NULL;
 }
 
 static int init_act_open(struct cxgbi_sock *csk)
@@ -1607,6 +1608,7 @@ static int init_act_open(struct cxgbi_sock *csk)
 	struct neighbour *n = NULL;
 	void *daddr;
 	unsigned int step;
+	unsigned int rxq_idx;
 	unsigned int size, size6;
 	unsigned int linkspeed;
 	unsigned int rcv_winf, snd_winf;
@@ -1685,7 +1687,9 @@ static int init_act_open(struct cxgbi_sock *csk)
 	step = lldi->ntxq / lldi->nchan;
 	csk->txq_idx = cxgb4_port_idx(ndev) * step;
 	step = lldi->nrxq / lldi->nchan;
-	csk->rss_qid = lldi->rxq_ids[cxgb4_port_idx(ndev) * step];
+	rxq_idx = (cxgb4_port_idx(ndev) * step) + (cdev->rxq_idx_cntr % step);
+	cdev->rxq_idx_cntr++;
+	csk->rss_qid = lldi->rxq_ids[rxq_idx];
 	linkspeed = ((struct port_info *)netdev_priv(ndev))->link_cfg.speed;
 	csk->snd_win = cxgb4i_snd_win;
 	csk->rcv_win = cxgb4i_rcv_win;
diff --git a/drivers/scsi/cxgbi/libcxgbi.c b/drivers/scsi/cxgbi/libcxgbi.c
index bd7d39ecbd24..e4c83b7c96a8 100644
--- a/drivers/scsi/cxgbi/libcxgbi.c
+++ b/drivers/scsi/cxgbi/libcxgbi.c
@@ -867,7 +867,8 @@ static void need_active_close(struct cxgbi_sock *csk)
 	log_debug(1 << CXGBI_DBG_SOCK, "csk 0x%p,%u,0x%lx,%u.\n",
 		csk, (csk)->state, (csk)->flags, (csk)->tid);
 	spin_lock_bh(&csk->lock);
-	dst_confirm(csk->dst);
+	if (csk->dst)
+		dst_confirm(csk->dst);
 	data_lost = skb_queue_len(&csk->receive_queue);
 	__skb_queue_purge(&csk->receive_queue);
 
@@ -882,7 +883,8 @@ static void need_active_close(struct cxgbi_sock *csk)
 	}
 
 	if (close_req) {
-		if (data_lost)
+		if (!cxgbi_sock_flag(csk, CTPF_LOGOUT_RSP_RCVD) ||
+		    data_lost)
 			csk->cdev->csk_send_abort_req(csk);
 		else
 			csk->cdev->csk_send_close_req(csk);
@@ -1186,9 +1188,10 @@ static int cxgbi_sock_send_pdus(struct cxgbi_sock *csk, struct sk_buff *skb)
 				cxgbi_ulp_extra_len(cxgbi_skcb_ulp_mode(skb));
 		skb = next;
 	}
-done:
+
 	if (likely(skb_queue_len(&csk->write_queue)))
 		cdev->csk_push_tx_frames(csk, 1);
+done:
 	spin_unlock_bh(&csk->lock);
 	return copied;
 
@@ -1568,9 +1571,12 @@ static inline int read_pdu_skb(struct iscsi_conn *conn,
 	}
 }
 
-static int skb_read_pdu_bhs(struct iscsi_conn *conn, struct sk_buff *skb)
+static int
+skb_read_pdu_bhs(struct cxgbi_sock *csk, struct iscsi_conn *conn,
+		 struct sk_buff *skb)
 {
 	struct iscsi_tcp_conn *tcp_conn = conn->dd_data;
+	int err;
 
 	log_debug(1 << CXGBI_DBG_PDU_RX,
 		"conn 0x%p, skb 0x%p, len %u, flag 0x%lx.\n",
@@ -1608,7 +1614,16 @@ static int skb_read_pdu_bhs(struct iscsi_conn *conn, struct sk_buff *skb)
 		}
 	}
 
-	return read_pdu_skb(conn, skb, 0, 0);
+	err = read_pdu_skb(conn, skb, 0, 0);
+	if (likely(err >= 0)) {
+		struct iscsi_hdr *hdr = (struct iscsi_hdr *)skb->data;
+		u8 opcode = hdr->opcode & ISCSI_OPCODE_MASK;
+
+		if (unlikely(opcode == ISCSI_OP_LOGOUT_RSP))
+			cxgbi_sock_set_flag(csk, CTPF_LOGOUT_RSP_RCVD);
+	}
+
+	return err;
 }
 
 static int skb_read_pdu_data(struct iscsi_conn *conn, struct sk_buff *lskb,
@@ -1713,7 +1728,7 @@ void cxgbi_conn_pdu_ready(struct cxgbi_sock *csk)
 			cxgbi_skcb_rx_pdulen(skb));
 
 		if (cxgbi_skcb_test_flag(skb, SKCBF_RX_COALESCED)) {
-			err = skb_read_pdu_bhs(conn, skb);
+			err = skb_read_pdu_bhs(csk, conn, skb);
 			if (err < 0) {
 				pr_err("coalesced bhs, csk 0x%p, skb 0x%p,%u, "
 					"f 0x%lx, plen %u.\n",
@@ -1731,7 +1746,7 @@ void cxgbi_conn_pdu_ready(struct cxgbi_sock *csk)
 					cxgbi_skcb_flags(skb),
 					cxgbi_skcb_rx_pdulen(skb));
 		} else {
-			err = skb_read_pdu_bhs(conn, skb);
+			err = skb_read_pdu_bhs(csk, conn, skb);
 			if (err < 0) {
 				pr_err("bhs, csk 0x%p, skb 0x%p,%u, "
 					"f 0x%lx, plen %u.\n",
@@ -1873,6 +1888,11 @@ int cxgbi_conn_alloc_pdu(struct iscsi_task *task, u8 opcode)
 	tcp_task->dd_data = tdata;
 	task->hdr = NULL;
 
+	if (tdata->skb) {
+		kfree_skb(tdata->skb);
+		tdata->skb = NULL;
+	}
+
 	if (SKB_MAX_HEAD(cdev->skb_tx_rsvd) > (512 * MAX_SKB_FRAGS) &&
 	    (opcode == ISCSI_OP_SCSI_DATA_OUT ||
 	     (opcode == ISCSI_OP_SCSI_CMD &&
@@ -1890,6 +1910,7 @@ int cxgbi_conn_alloc_pdu(struct iscsi_task *task, u8 opcode)
 		return -ENOMEM;
 	}
 
+	skb_get(tdata->skb);
 	skb_reserve(tdata->skb, cdev->skb_tx_rsvd);
 	task->hdr = (struct iscsi_hdr *)tdata->skb->data;
 	task->hdr_max = SKB_TX_ISCSI_PDU_HEADER_MAX; /* BHS + AHS */
@@ -2035,9 +2056,9 @@ int cxgbi_conn_xmit_pdu(struct iscsi_task *task)
 	unsigned int datalen;
 	int err;
 
-	if (!skb) {
+	if (!skb || cxgbi_skcb_test_flag(skb, SKCBF_TX_DONE)) {
 		log_debug(1 << CXGBI_DBG_ISCSI | 1 << CXGBI_DBG_PDU_TX,
-			"task 0x%p, skb NULL.\n", task);
+			"task 0x%p, skb 0x%p\n", task, skb);
 		return 0;
 	}
 
@@ -2050,7 +2071,6 @@ int cxgbi_conn_xmit_pdu(struct iscsi_task *task)
 	}
 
 	datalen = skb->data_len;
-	tdata->skb = NULL;
 
 	/* write ppod first if using ofldq to write ppod */
 	if (ttinfo->flags & CXGBI_PPOD_INFO_FLAG_VALID) {
@@ -2078,6 +2098,7 @@ int cxgbi_conn_xmit_pdu(struct iscsi_task *task)
 			pdulen += ISCSI_DIGEST_SIZE;
 
 		task->conn->txdata_octets += pdulen;
+		cxgbi_skcb_set_flag(skb, SKCBF_TX_DONE);
 		return 0;
 	}
 
@@ -2086,7 +2107,6 @@ int cxgbi_conn_xmit_pdu(struct iscsi_task *task)
 			"task 0x%p, skb 0x%p, len %u/%u, %d EAGAIN.\n",
 			task, skb, skb->len, skb->data_len, err);
 		/* reset skb to send when we are called again */
-		tdata->skb = skb;
 		return err;
 	}
 
@@ -2094,7 +2114,8 @@ int cxgbi_conn_xmit_pdu(struct iscsi_task *task)
 		"itt 0x%x, skb 0x%p, len %u/%u, xmit err %d.\n",
 		task->itt, skb, skb->len, skb->data_len, err);
 
-	kfree_skb(skb);
+	__kfree_skb(tdata->skb);
+	tdata->skb = NULL;
 
 	iscsi_conn_printk(KERN_ERR, task->conn, "xmit err %d.\n", err);
 	iscsi_conn_failure(task->conn, ISCSI_ERR_XMIT_FAILED);
@@ -2113,8 +2134,10 @@ void cxgbi_cleanup_task(struct iscsi_task *task)
 
 	tcp_task->dd_data = NULL;
 	/*  never reached the xmit task callout */
-	if (tdata->skb)
-		__kfree_skb(tdata->skb);
+	if (tdata->skb) {
+		kfree_skb(tdata->skb);
+		tdata->skb = NULL;
+	}
 
 	task_release_itt(task, task->hdr_itt);
 	memset(tdata, 0, sizeof(*tdata));
@@ -2714,6 +2737,9 @@ EXPORT_SYMBOL_GPL(cxgbi_attr_is_visible);
 static int __init libcxgbi_init_module(void)
 {
 	pr_info("%s", version);
+
+	BUILD_BUG_ON(FIELD_SIZEOF(struct sk_buff, cb) <
+		     sizeof(struct cxgbi_skb_cb));
 	return 0;
 }
 
diff --git a/drivers/scsi/cxgbi/libcxgbi.h b/drivers/scsi/cxgbi/libcxgbi.h
index 18e0ea83d361..31a5816c2e8d 100644
--- a/drivers/scsi/cxgbi/libcxgbi.h
+++ b/drivers/scsi/cxgbi/libcxgbi.h
@@ -187,6 +187,7 @@ enum cxgbi_sock_flags {
 	CTPF_HAS_ATID,		/* reserved atid */
 	CTPF_HAS_TID,		/* reserved hw tid */
 	CTPF_OFFLOAD_DOWN,	/* offload function off */
+	CTPF_LOGOUT_RSP_RCVD,   /* received logout response */
 };
 
 struct cxgbi_skb_rx_cb {
@@ -195,7 +196,8 @@ struct cxgbi_skb_rx_cb {
 };
 
 struct cxgbi_skb_tx_cb {
-	void *l2t;
+	void *handle;
+	void *arp_err_handler;
 	struct sk_buff *wr_next;
 };
 
@@ -203,6 +205,7 @@ enum cxgbi_skcb_flags {
 	SKCBF_TX_NEED_HDR,	/* packet needs a header */
 	SKCBF_TX_MEM_WRITE,     /* memory write */
 	SKCBF_TX_FLAG_COMPL,    /* wr completion flag */
+	SKCBF_TX_DONE,		/* skb tx done */
 	SKCBF_RX_COALESCED,	/* received whole pdu */
 	SKCBF_RX_HDR,		/* received pdu header */
 	SKCBF_RX_DATA,		/* received pdu payload */
@@ -215,13 +218,13 @@ enum cxgbi_skcb_flags {
 };
 
 struct cxgbi_skb_cb {
-	unsigned char ulp_mode;
-	unsigned long flags;
-	unsigned int seq;
 	union {
 		struct cxgbi_skb_rx_cb rx;
 		struct cxgbi_skb_tx_cb tx;
 	};
+	unsigned char ulp_mode;
+	unsigned long flags;
+	unsigned int seq;
 };
 
 #define CXGBI_SKB_CB(skb)	((struct cxgbi_skb_cb *)&((skb)->cb[0]))
@@ -374,11 +377,9 @@ static inline void cxgbi_sock_enqueue_wr(struct cxgbi_sock *csk,
 	cxgbi_skcb_tx_wr_next(skb) = NULL;
 	/*
 	 * We want to take an extra reference since both us and the driver
-	 * need to free the packet before it's really freed. We know there's
-	 * just one user currently so we use atomic_set rather than skb_get
-	 * to avoid the atomic op.
+	 * need to free the packet before it's really freed.
 	 */
-	atomic_set(&skb->users, 2);
+	skb_get(skb);
 
 	if (!csk->wr_pending_head)
 		csk->wr_pending_head = skb;
@@ -476,6 +477,7 @@ struct cxgbi_device {
 	unsigned int skb_rx_extra;	/* for msg coalesced mode */
 	unsigned int tx_max_size;
 	unsigned int rx_max_size;
+	unsigned int rxq_idx_cntr;
 	struct cxgbi_ports_map pmap;
 
 	void (*dev_ddp_cleanup)(struct cxgbi_device *);