Merge tag 'rdma-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland/infiniband

Pull infiniband updates from Roland Dreier:
 "First batch of InfiniBand/RDMA changes for the 3.7 merge window:
   - mlx4 IB support for SR-IOV
   - A couple of SRP initiator fixes
   - Batch of nes hardware driver fixes
   - Fix for long-standing use-after-free crash in IPoIB
   - Other miscellaneous fixes"

This merge also removes a new use of __cancel_delayed_work(), and
replaces it with the regular cancel_delayed_work() that is now irq-safe
thanks to the workqueue updates.

That said, I suspect the sequence in question should probably use
"mod_delayed_work()".  I just did the minimal "don't use deprecated
functions" fixup, though.

* tag 'rdma-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland/infiniband: (45 commits)
  IB/qib: Fix local access validation for user MRs
  mlx4_core: Disable SENSE_PORT for multifunction devices
  mlx4_core: Clean up enabling of SENSE_PORT for older (ConnectX-1/-2) HCAs
  mlx4_core: Stash PCI ID driver_data in mlx4_priv structure
  IB/srp: Avoid having aborted requests hang
  IB/srp: Fix use-after-free in srp_reset_req()
  IB/qib: Add a qib driver version
  RDMA/nes: Fix compilation error when nes_debug is enabled
  RDMA/nes: Print hardware resource type
  RDMA/nes: Fix for crash when TX checksum offload is off
  RDMA/nes: Cosmetic changes
  RDMA/nes: Fix for incorrect MSS when TSO is on
  RDMA/nes: Fix incorrect resolving of the loopback MAC address
  mlx4_core: Fix crash on uninitialized priv->cmd.slave_sem
  mlx4_core: Trivial cleanups to driver log messages
  mlx4_core: Trivial readability fix: "0X30" -> "0x30"
  IB/mlx4: Create paravirt contexts for VFs when master IB driver initializes
  mlx4: Modify proxy/tunnel QP mechanism so that guests do no calculations
  mlx4: Paravirtualize Node Guids for slaves
  mlx4: Activate SR-IOV mode for IB
  ...

drivers/infiniband/core/cache.c

@@ -167,6 +167,7 @@ int ib_find_cached_pkey(struct ib_device *device,
 	unsigned long flags;
 	int i;
 	int ret = -ENOENT;
+	int partial_ix = -1;
 
 	if (port_num < start_port(device) || port_num > end_port(device))
 		return -EINVAL;
@@ -179,9 +180,17 @@ int ib_find_cached_pkey(struct ib_device *device,
 
 	for (i = 0; i < cache->table_len; ++i)
 		if ((cache->table[i] & 0x7fff) == (pkey & 0x7fff)) {
+			if (cache->table[i] & 0x8000) {
 				*index = i;
 				ret = 0;
 				break;
+			} else
+				partial_ix = i;
+		}
+
+	if (ret && partial_ix >= 0) {
+		*index = partial_ix;
+		ret = 0;
 	}
 
 	read_unlock_irqrestore(&device->cache.lock, flags);
@@ -190,6 +199,38 @@ int ib_find_cached_pkey(struct ib_device *device,
 }
 EXPORT_SYMBOL(ib_find_cached_pkey);
 
+int ib_find_exact_cached_pkey(struct ib_device *device,
+			      u8                port_num,
+			      u16               pkey,
+			      u16              *index)
+{
+	struct ib_pkey_cache *cache;
+	unsigned long flags;
+	int i;
+	int ret = -ENOENT;
+
+	if (port_num < start_port(device) || port_num > end_port(device))
+		return -EINVAL;
+
+	read_lock_irqsave(&device->cache.lock, flags);
+
+	cache = device->cache.pkey_cache[port_num - start_port(device)];
+
+	*index = -1;
+
+	for (i = 0; i < cache->table_len; ++i)
+		if (cache->table[i] == pkey) {
+			*index = i;
+			ret = 0;
+			break;
+		}
+
+	read_unlock_irqrestore(&device->cache.lock, flags);
+
+	return ret;
+}
+EXPORT_SYMBOL(ib_find_exact_cached_pkey);
+
 int ib_get_cached_lmc(struct ib_device *device,
 		      u8                port_num,
 		      u8                *lmc)

drivers/infiniband/core/cma.c

@@ -3058,7 +3058,10 @@ static int cma_join_ib_multicast(struct rdma_id_private *id_priv,
 
 	if (id_priv->id.ps == RDMA_PS_IPOIB)
 		comp_mask |= IB_SA_MCMEMBER_REC_RATE |
-			     IB_SA_MCMEMBER_REC_RATE_SELECTOR;
+			     IB_SA_MCMEMBER_REC_RATE_SELECTOR |
+			     IB_SA_MCMEMBER_REC_MTU_SELECTOR |
+			     IB_SA_MCMEMBER_REC_MTU |
+			     IB_SA_MCMEMBER_REC_HOP_LIMIT;
 
 	mc->multicast.ib = ib_sa_join_multicast(&sa_client, id_priv->id.device,
 						id_priv->id.port_num, &rec,

drivers/infiniband/core/device.c

@@ -707,18 +707,28 @@ int ib_find_pkey(struct ib_device *device,
 {
 	int ret, i;
 	u16 tmp_pkey;
+	int partial_ix = -1;
 
 	for (i = 0; i < device->pkey_tbl_len[port_num - start_port(device)]; ++i) {
 		ret = ib_query_pkey(device, port_num, i, &tmp_pkey);
 		if (ret)
 			return ret;
-
 		if ((pkey & 0x7fff) == (tmp_pkey & 0x7fff)) {
+			/* if there is full-member pkey take it.*/
+			if (tmp_pkey & 0x8000) {
 				*index = i;
 				return 0;
 			}
+			if (partial_ix < 0)
+				partial_ix = i;
+		}
 	}
 
+	/*no full-member, if exists take the limited*/
+	if (partial_ix >= 0) {
+		*index = partial_ix;
+		return 0;
+	}
 	return -ENOENT;
 }
 EXPORT_SYMBOL(ib_find_pkey);

drivers/infiniband/core/ucm.c

@@ -397,7 +397,6 @@ static ssize_t ib_ucm_event(struct ib_ucm_file *file,
 	struct ib_ucm_event_get cmd;
 	struct ib_ucm_event *uevent;
 	int result = 0;
-	DEFINE_WAIT(wait);
 
 	if (out_len < sizeof(struct ib_ucm_event_resp))
 		return -ENOSPC;

drivers/infiniband/core/ucma.c

@@ -310,7 +310,6 @@ static ssize_t ucma_get_event(struct ucma_file *file, const char __user *inbuf,
 	struct rdma_ucm_get_event cmd;
 	struct ucma_event *uevent;
 	int ret = 0;
-	DEFINE_WAIT(wait);
 
 	if (out_len < sizeof uevent->resp)
 		return -ENOSPC;

drivers/infiniband/hw/cxgb4/qp.c

@@ -137,19 +137,25 @@ static int create_qp(struct c4iw_rdev *rdev, struct t4_wq *wq,
 		return -ENOMEM;
 
 	wq->rq.qid = c4iw_get_qpid(rdev, uctx);
-	if (!wq->rq.qid)
-		goto err1;
+	if (!wq->rq.qid) {
+		ret = -ENOMEM;
+		goto free_sq_qid;
+	}
 
 	if (!user) {
 		wq->sq.sw_sq = kzalloc(wq->sq.size * sizeof *wq->sq.sw_sq,
 				 GFP_KERNEL);
-		if (!wq->sq.sw_sq)
-			goto err2;
+		if (!wq->sq.sw_sq) {
+			ret = -ENOMEM;
+			goto free_rq_qid;
+		}
 
 		wq->rq.sw_rq = kzalloc(wq->rq.size * sizeof *wq->rq.sw_rq,
 				 GFP_KERNEL);
-		if (!wq->rq.sw_rq)
-			goto err3;
+		if (!wq->rq.sw_rq) {
+			ret = -ENOMEM;
+			goto free_sw_sq;
+		}
 	}
 
 	/*
@@ -157,15 +163,23 @@ static int create_qp(struct c4iw_rdev *rdev, struct t4_wq *wq,
 	 */
 	wq->rq.rqt_size = roundup_pow_of_two(wq->rq.size);
 	wq->rq.rqt_hwaddr = c4iw_rqtpool_alloc(rdev, wq->rq.rqt_size);
-	if (!wq->rq.rqt_hwaddr)
-		goto err4;
+	if (!wq->rq.rqt_hwaddr) {
+		ret = -ENOMEM;
+		goto free_sw_rq;
+	}
 
 	if (user) {
-		if (alloc_oc_sq(rdev, &wq->sq) && alloc_host_sq(rdev, &wq->sq))
-			goto err5;
+		ret = alloc_oc_sq(rdev, &wq->sq);
+		if (ret)
+			goto free_hwaddr;
+
+		ret = alloc_host_sq(rdev, &wq->sq);
+		if (ret)
+			goto free_sq;
 	} else
-		if (alloc_host_sq(rdev, &wq->sq))
-			goto err5;
+		ret = alloc_host_sq(rdev, &wq->sq);
+		if (ret)
+			goto free_hwaddr;
 	memset(wq->sq.queue, 0, wq->sq.memsize);
 	dma_unmap_addr_set(&wq->sq, mapping, wq->sq.dma_addr);
 
@@ -173,7 +187,7 @@ static int create_qp(struct c4iw_rdev *rdev, struct t4_wq *wq,
 					  wq->rq.memsize, &(wq->rq.dma_addr),
 					  GFP_KERNEL);
 	if (!wq->rq.queue)
-		goto err6;
+		goto free_sq;
 	PDBG("%s sq base va 0x%p pa 0x%llx rq base va 0x%p pa 0x%llx\n",
 		__func__, wq->sq.queue,
 		(unsigned long long)virt_to_phys(wq->sq.queue),
@@ -201,7 +215,7 @@ static int create_qp(struct c4iw_rdev *rdev, struct t4_wq *wq,
 	skb = alloc_skb(wr_len, GFP_KERNEL);
 	if (!skb) {
 		ret = -ENOMEM;
-		goto err7;
+		goto free_dma;
 	}
 	set_wr_txq(skb, CPL_PRIORITY_CONTROL, 0);
 
@@ -266,33 +280,33 @@ static int create_qp(struct c4iw_rdev *rdev, struct t4_wq *wq,
 
 	ret = c4iw_ofld_send(rdev, skb);
 	if (ret)
-		goto err7;
+		goto free_dma;
 	ret = c4iw_wait_for_reply(rdev, &wr_wait, 0, wq->sq.qid, __func__);
 	if (ret)
-		goto err7;
+		goto free_dma;
 
 	PDBG("%s sqid 0x%x rqid 0x%x kdb 0x%p squdb 0x%llx rqudb 0x%llx\n",
 	     __func__, wq->sq.qid, wq->rq.qid, wq->db,
 	     (unsigned long long)wq->sq.udb, (unsigned long long)wq->rq.udb);
 
 	return 0;
-err7:
+free_dma:
 	dma_free_coherent(&(rdev->lldi.pdev->dev),
 			  wq->rq.memsize, wq->rq.queue,
 			  dma_unmap_addr(&wq->rq, mapping));
-err6:
+free_sq:
 	dealloc_sq(rdev, &wq->sq);
-err5:
+free_hwaddr:
 	c4iw_rqtpool_free(rdev, wq->rq.rqt_hwaddr, wq->rq.rqt_size);
-err4:
+free_sw_rq:
 	kfree(wq->rq.sw_rq);
-err3:
+free_sw_sq:
 	kfree(wq->sq.sw_sq);
-err2:
+free_rq_qid:
 	c4iw_put_qpid(rdev, wq->rq.qid, uctx);
-err1:
+free_sq_qid:
 	c4iw_put_qpid(rdev, wq->sq.qid, uctx);
-	return -ENOMEM;
+	return ret;
 }
 
 static int build_immd(struct t4_sq *sq, struct fw_ri_immd *immdp,

drivers/infiniband/hw/mlx4/Makefile

 obj-$(CONFIG_MLX4_INFINIBAND)	+= mlx4_ib.o
 
-mlx4_ib-y :=	ah.o cq.o doorbell.o mad.o main.o mr.o qp.o srq.o
+mlx4_ib-y :=	ah.o cq.o doorbell.o mad.o main.o mr.o qp.o srq.o mcg.o cm.o alias_GUID.o sysfs.o

drivers/infiniband/hw/mlx4/cq.c

@@ -547,6 +547,26 @@ static int mlx4_ib_ipoib_csum_ok(__be16 status, __be16 checksum)
 		checksum == cpu_to_be16(0xffff);
 }
 
+static int use_tunnel_data(struct mlx4_ib_qp *qp, struct mlx4_ib_cq *cq, struct ib_wc *wc,
+			   unsigned tail, struct mlx4_cqe *cqe)
+{
+	struct mlx4_ib_proxy_sqp_hdr *hdr;
+
+	ib_dma_sync_single_for_cpu(qp->ibqp.device,
+				   qp->sqp_proxy_rcv[tail].map,
+				   sizeof (struct mlx4_ib_proxy_sqp_hdr),
+				   DMA_FROM_DEVICE);
+	hdr = (struct mlx4_ib_proxy_sqp_hdr *) (qp->sqp_proxy_rcv[tail].addr);
+	wc->pkey_index	= be16_to_cpu(hdr->tun.pkey_index);
+	wc->slid	= be16_to_cpu(hdr->tun.slid_mac_47_32);
+	wc->sl		= (u8) (be16_to_cpu(hdr->tun.sl_vid) >> 12);
+	wc->src_qp	= be32_to_cpu(hdr->tun.flags_src_qp) & 0xFFFFFF;
+	wc->wc_flags   |= (hdr->tun.g_ml_path & 0x80) ? (IB_WC_GRH) : 0;
+	wc->dlid_path_bits = 0;
+
+	return 0;
+}
+
 static int mlx4_ib_poll_one(struct mlx4_ib_cq *cq,
 			    struct mlx4_ib_qp **cur_qp,
 			    struct ib_wc *wc)
@@ -559,6 +579,7 @@ static int mlx4_ib_poll_one(struct mlx4_ib_cq *cq,
 	int is_error;
 	u32 g_mlpath_rqpn;
 	u16 wqe_ctr;
+	unsigned tail = 0;
 
 repoll:
 	cqe = next_cqe_sw(cq);
@@ -634,7 +655,8 @@ static int mlx4_ib_poll_one(struct mlx4_ib_cq *cq,
 		mlx4_ib_free_srq_wqe(srq, wqe_ctr);
 	} else {
 		wq	  = &(*cur_qp)->rq;
-		wc->wr_id = wq->wrid[wq->tail & (wq->wqe_cnt - 1)];
+		tail	  = wq->tail & (wq->wqe_cnt - 1);
+		wc->wr_id = wq->wrid[tail];
 		++wq->tail;
 	}
 
@@ -717,6 +739,13 @@ static int mlx4_ib_poll_one(struct mlx4_ib_cq *cq,
 			break;
 		}
 
+		if (mlx4_is_mfunc(to_mdev(cq->ibcq.device)->dev)) {
+			if ((*cur_qp)->mlx4_ib_qp_type &
+			    (MLX4_IB_QPT_PROXY_SMI_OWNER |
+			     MLX4_IB_QPT_PROXY_SMI | MLX4_IB_QPT_PROXY_GSI))
+				return use_tunnel_data(*cur_qp, cq, wc, tail, cqe);
+		}
+
 		wc->slid	   = be16_to_cpu(cqe->rlid);
 		g_mlpath_rqpn	   = be32_to_cpu(cqe->g_mlpath_rqpn);
 		wc->src_qp	   = g_mlpath_rqpn & 0xffffff;

drivers/infiniband/hw/nes/nes.h

@@ -399,11 +399,20 @@ static inline void nes_write8(void __iomem *addr, u8 val)
 	writeb(val, addr);
 }
 
-
+enum nes_resource {
+	NES_RESOURCE_MW = 1,
+	NES_RESOURCE_FAST_MR,
+	NES_RESOURCE_PHYS_MR,
+	NES_RESOURCE_USER_MR,
+	NES_RESOURCE_PD,
+	NES_RESOURCE_QP,
+	NES_RESOURCE_CQ,
+	NES_RESOURCE_ARP
+};
 
 static inline int nes_alloc_resource(struct nes_adapter *nesadapter,
 		unsigned long *resource_array, u32 max_resources,
-		u32 *req_resource_num, u32 *next)
+		u32 *req_resource_num, u32 *next, enum nes_resource resource_type)
 {
 	unsigned long flags;
 	u32 resource_num;
@@ -414,7 +423,7 @@ static inline int nes_alloc_resource(struct nes_adapter *nesadapter,
 	if (resource_num >= max_resources) {
 		resource_num = find_first_zero_bit(resource_array, max_resources);
 		if (resource_num >= max_resources) {
-			printk(KERN_ERR PFX "%s: No available resourcess.\n", __func__);
+			printk(KERN_ERR PFX "%s: No available resources [type=%u].\n", __func__, resource_type);
 			spin_unlock_irqrestore(&nesadapter->resource_lock, flags);
 			return -EMFILE;
 		}

drivers/infiniband/hw/nes/nes_cm.c

@@ -430,6 +430,8 @@ static void form_cm_frame(struct sk_buff *skb,
 	buf += sizeof(*tcph);
 
 	skb->ip_summed = CHECKSUM_PARTIAL;
+	if (!(cm_node->netdev->features & NETIF_F_IP_CSUM))
+		skb->ip_summed = CHECKSUM_NONE;
 	skb->protocol = htons(0x800);
 	skb->data_len = 0;
 	skb->mac_len = ETH_HLEN;
@@ -1356,7 +1358,7 @@ static int nes_addr_resolve_neigh(struct nes_vnic *nesvnic, u32 dst_ip, int arpi
 	else
 		netdev = nesvnic->netdev;
 
-	neigh = dst_neigh_lookup(&rt->dst, &dst_ip);
+	neigh = neigh_lookup(&arp_tbl, &rt->rt_gateway, netdev);
 
 	rcu_read_lock();
 	if (neigh) {
@@ -1465,12 +1467,8 @@ static struct nes_cm_node *make_cm_node(struct nes_cm_core *cm_core,
 	cm_node->loopbackpartner = NULL;
 
 	/* get the mac addr for the remote node */
-	if (ipv4_is_loopback(htonl(cm_node->rem_addr))) {
-		arpindex = nes_arp_table(nesdev, ntohl(nesvnic->local_ipaddr), NULL, NES_ARP_RESOLVE);
-	} else {
 	oldarpindex = nes_arp_table(nesdev, cm_node->rem_addr, NULL, NES_ARP_RESOLVE);
 	arpindex = nes_addr_resolve_neigh(nesvnic, cm_info->rem_addr, oldarpindex);
-	}
 	if (arpindex < 0) {
 		kfree(cm_node);
 		return NULL;
@@ -3153,10 +3151,6 @@ int nes_accept(struct iw_cm_id *cm_id, struct iw_cm_conn_param *conn_param)
 	nesqp->nesqp_context->tcpPorts[1] =
 		cpu_to_le16(ntohs(cm_id->remote_addr.sin_port));
 
-	if (ipv4_is_loopback(cm_id->remote_addr.sin_addr.s_addr))
-		nesqp->nesqp_context->ip0 =
-			cpu_to_le32(ntohl(nesvnic->local_ipaddr));
-	else
 	nesqp->nesqp_context->ip0 =
 			cpu_to_le32(ntohl(cm_id->remote_addr.sin_addr.s_addr));
 
@@ -3182,9 +3176,6 @@ int nes_accept(struct iw_cm_id *cm_id, struct iw_cm_conn_param *conn_param)
 	memset(&nes_quad, 0, sizeof(nes_quad));
 	nes_quad.DstIpAdrIndex =
 		cpu_to_le32((u32)PCI_FUNC(nesdev->pcidev->devfn) << 24);
-	if (ipv4_is_loopback(cm_id->remote_addr.sin_addr.s_addr))
-		nes_quad.SrcIpadr = nesvnic->local_ipaddr;
-	else
 	nes_quad.SrcIpadr = cm_id->remote_addr.sin_addr.s_addr;
 	nes_quad.TcpPorts[0] = cm_id->remote_addr.sin_port;
 	nes_quad.TcpPorts[1] = cm_id->local_addr.sin_port;
@@ -3538,10 +3529,6 @@ static void cm_event_connected(struct nes_cm_event *event)
 		cpu_to_le16(ntohs(cm_id->local_addr.sin_port));
 	nesqp->nesqp_context->tcpPorts[1] =
 		cpu_to_le16(ntohs(cm_id->remote_addr.sin_port));
-	if (ipv4_is_loopback(cm_id->remote_addr.sin_addr.s_addr))
-		nesqp->nesqp_context->ip0 =
-			cpu_to_le32(ntohl(nesvnic->local_ipaddr));
-	else
 	nesqp->nesqp_context->ip0 =
 			cpu_to_le32(ntohl(cm_id->remote_addr.sin_addr.s_addr));
 
@@ -3571,9 +3558,6 @@ static void cm_event_connected(struct nes_cm_event *event)
 
 	nes_quad.DstIpAdrIndex =
 		cpu_to_le32((u32)PCI_FUNC(nesdev->pcidev->devfn) << 24);
-	if (ipv4_is_loopback(cm_id->remote_addr.sin_addr.s_addr))
-		nes_quad.SrcIpadr = nesvnic->local_ipaddr;
-	else
 	nes_quad.SrcIpadr = cm_id->remote_addr.sin_addr.s_addr;
 	nes_quad.TcpPorts[0] = cm_id->remote_addr.sin_port;
 	nes_quad.TcpPorts[1] = cm_id->local_addr.sin_port;

drivers/infiniband/hw/nes/nes_hw.c

@@ -3575,10 +3575,10 @@ static void nes_process_iwarp_aeqe(struct nes_device *nesdev,
 	tcp_state = (aeq_info & NES_AEQE_TCP_STATE_MASK) >> NES_AEQE_TCP_STATE_SHIFT;
 	iwarp_state = (aeq_info & NES_AEQE_IWARP_STATE_MASK) >> NES_AEQE_IWARP_STATE_SHIFT;
 	nes_debug(NES_DBG_AEQ, "aeid = 0x%04X, qp-cq id = %d, aeqe = %p,"
-			" Tcp state = %s, iWARP state = %s\n",
+			" Tcp state = %d, iWARP state = %d\n",
 			async_event_id,
 			le32_to_cpu(aeqe->aeqe_words[NES_AEQE_COMP_QP_CQ_ID_IDX]), aeqe,
-			nes_tcp_state_str[tcp_state], nes_iwarp_state_str[iwarp_state]);
+			tcp_state, iwarp_state);
 
 	aeqe_cq_id = le32_to_cpu(aeqe->aeqe_words[NES_AEQE_COMP_QP_CQ_ID_IDX]);
 	if (aeq_info & NES_AEQE_QP) {