Merge branch 'k.o/for-4.10-rc' into HEAD

7 files changed, 159 insertions, 84 deletions

diff --git a/drivers/infiniband/hw/cxgb4/cm.c b/drivers/infiniband/hw/cxgb4/cm.c
index 9749c52f8729..a33f37998e63 100644
--- a/drivers/infiniband/hw/cxgb4/cm.c
+++ b/drivers/infiniband/hw/cxgb4/cm.c
@@ -1804,20 +1804,21 @@ static int rx_data(struct c4iw_dev *dev, struct sk_buff *skb)
 	skb_trim(skb, dlen);
 	mutex_lock(&ep->com.mutex);
 
-	/* update RX credits */
-	update_rx_credits(ep, dlen);
-
 	switch (ep->com.state) {
 	case MPA_REQ_SENT:
+		update_rx_credits(ep, dlen);
 		ep->rcv_seq += dlen;
 		disconnect = process_mpa_reply(ep, skb);
 		break;
 	case MPA_REQ_WAIT:
+		update_rx_credits(ep, dlen);
 		ep->rcv_seq += dlen;
 		disconnect = process_mpa_request(ep, skb);
 		break;
 	case FPDU_MODE: {
 		struct c4iw_qp_attributes attrs;
+
+		update_rx_credits(ep, dlen);
 		BUG_ON(!ep->com.qp);
 		if (status)
 			pr_err("%s Unexpected streaming data." \
diff --git a/drivers/infiniband/hw/cxgb4/cq.c b/drivers/infiniband/hw/cxgb4/cq.c
index 19c6477af19f..bec82a600d77 100644
--- a/drivers/infiniband/hw/cxgb4/cq.c
+++ b/drivers/infiniband/hw/cxgb4/cq.c
@@ -505,6 +505,15 @@ static int poll_cq(struct t4_wq *wq, struct t4_cq *cq, struct t4_cqe *cqe,
 	}
 
 	/*
+	 * Special cqe for drain WR completions...
+	 */
+	if (CQE_OPCODE(hw_cqe) == C4IW_DRAIN_OPCODE) {
+		*cookie = CQE_DRAIN_COOKIE(hw_cqe);
+		*cqe = *hw_cqe;
+		goto skip_cqe;
+	}
+
+	/*
 	 * Gotta tweak READ completions:
 	 *	1) the cqe doesn't contain the sq_wptr from the wr.
 	 *	2) opcode not reflected from the wr.
@@ -753,6 +762,9 @@ static int c4iw_poll_cq_one(struct c4iw_cq *chp, struct ib_wc *wc)
 				c4iw_invalidate_mr(qhp->rhp,
 						   CQE_WRID_FR_STAG(&cqe));
 			break;
+		case C4IW_DRAIN_OPCODE:
+			wc->opcode = IB_WC_SEND;
+			break;
 		default:
 			printk(KERN_ERR MOD "Unexpected opcode %d "
 			       "in the CQE received for QPID=0x%0x\n",
@@ -817,15 +829,8 @@ static int c4iw_poll_cq_one(struct c4iw_cq *chp, struct ib_wc *wc)
 		}
 	}
 out:
-	if (wq) {
-		if (unlikely(qhp->attr.state != C4IW_QP_STATE_RTS)) {
-			if (t4_sq_empty(wq))
-				complete(&qhp->sq_drained);
-			if (t4_rq_empty(wq))
-				complete(&qhp->rq_drained);
-		}
+	if (wq)
 		spin_unlock(&qhp->lock);
-	}
 	return ret;
 }
 
diff --git a/drivers/infiniband/hw/cxgb4/device.c b/drivers/infiniband/hw/cxgb4/device.c
index c4b344e1f7e7..4e4f1a732b01 100644
--- a/drivers/infiniband/hw/cxgb4/device.c
+++ b/drivers/infiniband/hw/cxgb4/device.c
@@ -881,9 +881,17 @@ static int c4iw_rdev_open(struct c4iw_rdev *rdev)
 		}
 	}
 
+	rdev->free_workq = create_singlethread_workqueue("iw_cxgb4_free");
+	if (!rdev->free_workq) {
+		err = -ENOMEM;
+		goto err_free_status_page;
+	}
+
 	rdev->status_page->db_off = 0;
 
 	return 0;
+err_free_status_page:
+	free_page((unsigned long)rdev->status_page);
 destroy_ocqp_pool:
 	c4iw_ocqp_pool_destroy(rdev);
 destroy_rqtpool:
@@ -897,6 +905,7 @@ destroy_resource:
 
 static void c4iw_rdev_close(struct c4iw_rdev *rdev)
 {
+	destroy_workqueue(rdev->free_workq);
 	kfree(rdev->wr_log);
 	free_page((unsigned long)rdev->status_page);
 	c4iw_pblpool_destroy(rdev);
diff --git a/drivers/infiniband/hw/cxgb4/iw_cxgb4.h b/drivers/infiniband/hw/cxgb4/iw_cxgb4.h
index 4788e1a46fde..8cd4d054a87e 100644
--- a/drivers/infiniband/hw/cxgb4/iw_cxgb4.h
+++ b/drivers/infiniband/hw/cxgb4/iw_cxgb4.h
@@ -45,6 +45,7 @@
 #include <linux/kref.h>
 #include <linux/timer.h>
 #include <linux/io.h>
+#include <linux/workqueue.h>
 
 #include <asm/byteorder.h>
 
@@ -107,6 +108,7 @@ struct c4iw_dev_ucontext {
 	struct list_head qpids;
 	struct list_head cqids;
 	struct mutex lock;
+	struct kref kref;
 };
 
 enum c4iw_rdev_flags {
@@ -183,6 +185,7 @@ struct c4iw_rdev {
 	atomic_t wr_log_idx;
 	struct wr_log_entry *wr_log;
 	int wr_log_size;
+	struct workqueue_struct *free_workq;
 };
 
 static inline int c4iw_fatal_error(struct c4iw_rdev *rdev)
@@ -480,8 +483,8 @@ struct c4iw_qp {
 	wait_queue_head_t wait;
 	struct timer_list timer;
 	int sq_sig_all;
-	struct completion rq_drained;
-	struct completion sq_drained;
+	struct work_struct free_work;
+	struct c4iw_ucontext *ucontext;
 };
 
 static inline struct c4iw_qp *to_c4iw_qp(struct ib_qp *ibqp)
@@ -495,6 +498,7 @@ struct c4iw_ucontext {
 	u32 key;
 	spinlock_t mmap_lock;
 	struct list_head mmaps;
+	struct kref kref;
 };
 
 static inline struct c4iw_ucontext *to_c4iw_ucontext(struct ib_ucontext *c)
@@ -502,6 +506,18 @@ static inline struct c4iw_ucontext *to_c4iw_ucontext(struct ib_ucontext *c)
 	return container_of(c, struct c4iw_ucontext, ibucontext);
 }
 
+void _c4iw_free_ucontext(struct kref *kref);
+
+static inline void c4iw_put_ucontext(struct c4iw_ucontext *ucontext)
+{
+	kref_put(&ucontext->kref, _c4iw_free_ucontext);
+}
+
+static inline void c4iw_get_ucontext(struct c4iw_ucontext *ucontext)
+{
+	kref_get(&ucontext->kref);
+}
+
 struct c4iw_mm_entry {
 	struct list_head entry;
 	u64 addr;
@@ -615,6 +631,8 @@ static inline int to_ib_qp_state(int c4iw_qp_state)
 	return IB_QPS_ERR;
 }
 
+#define C4IW_DRAIN_OPCODE FW_RI_SGE_EC_CR_RETURN
+
 static inline u32 c4iw_ib_to_tpt_access(int a)
 {
 	return (a & IB_ACCESS_REMOTE_WRITE ? FW_RI_MEM_ACCESS_REM_WRITE : 0) |
@@ -997,8 +1015,6 @@ extern int c4iw_wr_log;
 extern int db_fc_threshold;
 extern int db_coalescing_threshold;
 extern int use_dsgl;
-void c4iw_drain_rq(struct ib_qp *qp);
-void c4iw_drain_sq(struct ib_qp *qp);
 void c4iw_invalidate_mr(struct c4iw_dev *rhp, u32 rkey);
 
 #endif
diff --git a/drivers/infiniband/hw/cxgb4/provider.c b/drivers/infiniband/hw/cxgb4/provider.c
index 49b51b7e0fd7..3345e1c312f7 100644
--- a/drivers/infiniband/hw/cxgb4/provider.c
+++ b/drivers/infiniband/hw/cxgb4/provider.c
@@ -93,17 +93,28 @@ static int c4iw_process_mad(struct ib_device *ibdev, int mad_flags,
 	return -ENOSYS;
 }
 
-static int c4iw_dealloc_ucontext(struct ib_ucontext *context)
+void _c4iw_free_ucontext(struct kref *kref)
 {
-	struct c4iw_dev *rhp = to_c4iw_dev(context->device);
-	struct c4iw_ucontext *ucontext = to_c4iw_ucontext(context);
+	struct c4iw_ucontext *ucontext;
+	struct c4iw_dev *rhp;
 	struct c4iw_mm_entry *mm, *tmp;
 
-	PDBG("%s context %p\n", __func__, context);
+	ucontext = container_of(kref, struct c4iw_ucontext, kref);
+	rhp = to_c4iw_dev(ucontext->ibucontext.device);
+
+	PDBG("%s ucontext %p\n", __func__, ucontext);
 	list_for_each_entry_safe(mm, tmp, &ucontext->mmaps, entry)
 		kfree(mm);
 	c4iw_release_dev_ucontext(&rhp->rdev, &ucontext->uctx);
 	kfree(ucontext);
+}
+
+static int c4iw_dealloc_ucontext(struct ib_ucontext *context)
+{
+	struct c4iw_ucontext *ucontext = to_c4iw_ucontext(context);
+
+	PDBG("%s context %p\n", __func__, context);
+	c4iw_put_ucontext(ucontext);
 	return 0;
 }
 
@@ -127,6 +138,7 @@ static struct ib_ucontext *c4iw_alloc_ucontext(struct ib_device *ibdev,
 	c4iw_init_dev_ucontext(&rhp->rdev, &context->uctx);
 	INIT_LIST_HEAD(&context->mmaps);
 	spin_lock_init(&context->mmap_lock);
+	kref_init(&context->kref);
 
 	if (udata->outlen < sizeof(uresp) - sizeof(uresp.reserved)) {
 		if (!warned++)
@@ -361,16 +373,7 @@ static int c4iw_query_port(struct ib_device *ibdev, u8 port,
 
 	memset(props, 0, sizeof(struct ib_port_attr));
 	props->max_mtu = IB_MTU_4096;
-	if (netdev->mtu >= 4096)
-		props->active_mtu = IB_MTU_4096;
-	else if (netdev->mtu >= 2048)
-		props->active_mtu = IB_MTU_2048;
-	else if (netdev->mtu >= 1024)
-		props->active_mtu = IB_MTU_1024;
-	else if (netdev->mtu >= 512)
-		props->active_mtu = IB_MTU_512;
-	else
-		props->active_mtu = IB_MTU_256;
+	props->active_mtu = ib_mtu_int_to_enum(netdev->mtu);
 
 	if (!netif_carrier_ok(netdev))
 		props->state = IB_PORT_DOWN;
@@ -607,8 +610,6 @@ int c4iw_register_device(struct c4iw_dev *dev)
 	dev->ibdev.uverbs_abi_ver = C4IW_UVERBS_ABI_VERSION;
 	dev->ibdev.get_port_immutable = c4iw_port_immutable;
 	dev->ibdev.get_dev_fw_str = get_dev_fw_str;
-	dev->ibdev.drain_sq = c4iw_drain_sq;
-	dev->ibdev.drain_rq = c4iw_drain_rq;
 
 	dev->ibdev.iwcm = kmalloc(sizeof(struct iw_cm_verbs), GFP_KERNEL);
 	if (!dev->ibdev.iwcm)
diff --git a/drivers/infiniband/hw/cxgb4/qp.c b/drivers/infiniband/hw/cxgb4/qp.c
index cda5542e13a2..04c1c382dedb 100644
--- a/drivers/infiniband/hw/cxgb4/qp.c
+++ b/drivers/infiniband/hw/cxgb4/qp.c
@@ -715,13 +715,32 @@ static int build_inv_stag(union t4_wr *wqe, struct ib_send_wr *wr, u8 *len16)
 	return 0;
 }
 
-static void _free_qp(struct kref *kref)
+static void free_qp_work(struct work_struct *work)
+{
+	struct c4iw_ucontext *ucontext;
+	struct c4iw_qp *qhp;
+	struct c4iw_dev *rhp;
+
+	qhp = container_of(work, struct c4iw_qp, free_work);
+	ucontext = qhp->ucontext;
+	rhp = qhp->rhp;
+
+	PDBG("%s qhp %p ucontext %p\n", __func__, qhp, ucontext);
+	destroy_qp(&rhp->rdev, &qhp->wq,
+		   ucontext ? &ucontext->uctx : &rhp->rdev.uctx);
+
+	if (ucontext)
+		c4iw_put_ucontext(ucontext);
+	kfree(qhp);
+}
+
+static void queue_qp_free(struct kref *kref)
 {
 	struct c4iw_qp *qhp;
 
 	qhp = container_of(kref, struct c4iw_qp, kref);
 	PDBG("%s qhp %p\n", __func__, qhp);
-	kfree(qhp);
+	queue_work(qhp->rhp->rdev.free_workq, &qhp->free_work);
 }
 
 void c4iw_qp_add_ref(struct ib_qp *qp)
@@ -733,7 +752,7 @@ void c4iw_qp_add_ref(struct ib_qp *qp)
 void c4iw_qp_rem_ref(struct ib_qp *qp)
 {
 	PDBG("%s ib_qp %p\n", __func__, qp);
-	kref_put(&to_c4iw_qp(qp)->kref, _free_qp);
+	kref_put(&to_c4iw_qp(qp)->kref, queue_qp_free);
 }
 
 static void add_to_fc_list(struct list_head *head, struct list_head *entry)
@@ -776,6 +795,64 @@ static int ring_kernel_rq_db(struct c4iw_qp *qhp, u16 inc)
 	return 0;
 }
 
+static void complete_sq_drain_wr(struct c4iw_qp *qhp, struct ib_send_wr *wr)
+{
+	struct t4_cqe cqe = {};
+	struct c4iw_cq *schp;
+	unsigned long flag;
+	struct t4_cq *cq;
+
+	schp = to_c4iw_cq(qhp->ibqp.send_cq);
+	cq = &schp->cq;
+
+	cqe.u.drain_cookie = wr->wr_id;
+	cqe.header = cpu_to_be32(CQE_STATUS_V(T4_ERR_SWFLUSH) |
+				 CQE_OPCODE_V(C4IW_DRAIN_OPCODE) |
+				 CQE_TYPE_V(1) |
+				 CQE_SWCQE_V(1) |
+				 CQE_QPID_V(qhp->wq.sq.qid));
+
+	spin_lock_irqsave(&schp->lock, flag);
+	cqe.bits_type_ts = cpu_to_be64(CQE_GENBIT_V((u64)cq->gen));
+	cq->sw_queue[cq->sw_pidx] = cqe;
+	t4_swcq_produce(cq);
+	spin_unlock_irqrestore(&schp->lock, flag);
+
+	spin_lock_irqsave(&schp->comp_handler_lock, flag);
+	(*schp->ibcq.comp_handler)(&schp->ibcq,
+				   schp->ibcq.cq_context);
+	spin_unlock_irqrestore(&schp->comp_handler_lock, flag);
+}
+
+static void complete_rq_drain_wr(struct c4iw_qp *qhp, struct ib_recv_wr *wr)
+{
+	struct t4_cqe cqe = {};
+	struct c4iw_cq *rchp;
+	unsigned long flag;
+	struct t4_cq *cq;
+
+	rchp = to_c4iw_cq(qhp->ibqp.recv_cq);
+	cq = &rchp->cq;
+
+	cqe.u.drain_cookie = wr->wr_id;
+	cqe.header = cpu_to_be32(CQE_STATUS_V(T4_ERR_SWFLUSH) |
+				 CQE_OPCODE_V(C4IW_DRAIN_OPCODE) |
+				 CQE_TYPE_V(0) |
+				 CQE_SWCQE_V(1) |
+				 CQE_QPID_V(qhp->wq.sq.qid));
+
+	spin_lock_irqsave(&rchp->lock, flag);
+	cqe.bits_type_ts = cpu_to_be64(CQE_GENBIT_V((u64)cq->gen));
+	cq->sw_queue[cq->sw_pidx] = cqe;
+	t4_swcq_produce(cq);
+	spin_unlock_irqrestore(&rchp->lock, flag);
+
+	spin_lock_irqsave(&rchp->comp_handler_lock, flag);
+	(*rchp->ibcq.comp_handler)(&rchp->ibcq,
+				   rchp->ibcq.cq_context);
+	spin_unlock_irqrestore(&rchp->comp_handler_lock, flag);
+}
+
 int c4iw_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr,
 		   struct ib_send_wr **bad_wr)
 {
@@ -794,8 +871,8 @@ int c4iw_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr,
 	spin_lock_irqsave(&qhp->lock, flag);
 	if (t4_wq_in_error(&qhp->wq)) {
 		spin_unlock_irqrestore(&qhp->lock, flag);
-		*bad_wr = wr;
-		return -EINVAL;
+		complete_sq_drain_wr(qhp, wr);
+		return err;
 	}
 	num_wrs = t4_sq_avail(&qhp->wq);
 	if (num_wrs == 0) {
@@ -937,8 +1014,8 @@ int c4iw_post_receive(struct ib_qp *ibqp, struct ib_recv_wr *wr,
 	spin_lock_irqsave(&qhp->lock, flag);
 	if (t4_wq_in_error(&qhp->wq)) {
 		spin_unlock_irqrestore(&qhp->lock, flag);
-		*bad_wr = wr;
-		return -EINVAL;
+		complete_rq_drain_wr(qhp, wr);
+		return err;
 	}
 	num_wrs = t4_rq_avail(&qhp->wq);
 	if (num_wrs == 0) {
@@ -1550,7 +1627,12 @@ int c4iw_modify_qp(struct c4iw_dev *rhp, struct c4iw_qp *qhp,
 		}
 		break;
 	case C4IW_QP_STATE_CLOSING:
-		if (!internal) {
+
+		/*
+		 * Allow kernel users to move to ERROR for qp draining.
+		 */
+		if (!internal && (qhp->ibqp.uobject || attrs->next_state !=
+				  C4IW_QP_STATE_ERROR)) {
 			ret = -EINVAL;
 			goto out;
 		}
@@ -1643,7 +1725,6 @@ int c4iw_destroy_qp(struct ib_qp *ib_qp)
 	struct c4iw_dev *rhp;
 	struct c4iw_qp *qhp;
 	struct c4iw_qp_attributes attrs;
-	struct c4iw_ucontext *ucontext;
 
 	qhp = to_c4iw_qp(ib_qp);
 	rhp = qhp->rhp;
@@ -1663,11 +1744,6 @@ int c4iw_destroy_qp(struct ib_qp *ib_qp)
 	spin_unlock_irq(&rhp->lock);
 	free_ird(rhp, qhp->attr.max_ird);
 
-	ucontext = ib_qp->uobject ?
-		   to_c4iw_ucontext(ib_qp->uobject->context) : NULL;
-	destroy_qp(&rhp->rdev, &qhp->wq,
-		   ucontext ? &ucontext->uctx : &rhp->rdev.uctx);
-
 	c4iw_qp_rem_ref(ib_qp);
 
 	PDBG("%s ib_qp %p qpid 0x%0x\n", __func__, ib_qp, qhp->wq.sq.qid);
@@ -1763,11 +1839,10 @@ struct ib_qp *c4iw_create_qp(struct ib_pd *pd, struct ib_qp_init_attr *attrs,
 	qhp->attr.max_ird = 0;
 	qhp->sq_sig_all = attrs->sq_sig_type == IB_SIGNAL_ALL_WR;
 	spin_lock_init(&qhp->lock);
-	init_completion(&qhp->sq_drained);
-	init_completion(&qhp->rq_drained);
 	mutex_init(&qhp->mutex);
 	init_waitqueue_head(&qhp->wait);
 	kref_init(&qhp->kref);
+	INIT_WORK(&qhp->free_work, free_qp_work);
 
 	ret = insert_handle(rhp, &rhp->qpidr, qhp, qhp->wq.sq.qid);
 	if (ret)
@@ -1854,6 +1929,9 @@ struct ib_qp *c4iw_create_qp(struct ib_pd *pd, struct ib_qp_init_attr *attrs,
 			ma_sync_key_mm->len = PAGE_SIZE;
 			insert_mmap(ucontext, ma_sync_key_mm);
 		}
+
+		c4iw_get_ucontext(ucontext);
+		qhp->ucontext = ucontext;
 	}
 	qhp->ibqp.qp_num = qhp->wq.sq.qid;
 	init_timer(&(qhp->timer));
@@ -1958,40 +2036,3 @@ int c4iw_ib_query_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
 	init_attr->sq_sig_type = qhp->sq_sig_all ? IB_SIGNAL_ALL_WR : 0;
 	return 0;
 }
-
-static void move_qp_to_err(struct c4iw_qp *qp)
-{
-	struct c4iw_qp_attributes attrs = { .next_state = C4IW_QP_STATE_ERROR };
-
-	(void)c4iw_modify_qp(qp->rhp, qp, C4IW_QP_ATTR_NEXT_STATE, &attrs, 1);
-}
-
-void c4iw_drain_sq(struct ib_qp *ibqp)
-{
-	struct c4iw_qp *qp = to_c4iw_qp(ibqp);
-	unsigned long flag;
-	bool need_to_wait;
-
-	move_qp_to_err(qp);
-	spin_lock_irqsave(&qp->lock, flag);
-	need_to_wait = !t4_sq_empty(&qp->wq);
-	spin_unlock_irqrestore(&qp->lock, flag);
-
-	if (need_to_wait)
-		wait_for_completion(&qp->sq_drained);
-}
-
-void c4iw_drain_rq(struct ib_qp *ibqp)
-{
-	struct c4iw_qp *qp = to_c4iw_qp(ibqp);
-	unsigned long flag;
-	bool need_to_wait;
-
-	move_qp_to_err(qp);
-	spin_lock_irqsave(&qp->lock, flag);
-	need_to_wait = !t4_rq_empty(&qp->wq);
-	spin_unlock_irqrestore(&qp->lock, flag);
-
-	if (need_to_wait)
-		wait_for_completion(&qp->rq_drained);
-}
diff --git a/drivers/infiniband/hw/cxgb4/t4.h b/drivers/infiniband/hw/cxgb4/t4.h
index 862381aa83c8..640d22148a3e 100644
--- a/drivers/infiniband/hw/cxgb4/t4.h
+++ b/drivers/infiniband/hw/cxgb4/t4.h
@@ -179,6 +179,7 @@ struct t4_cqe {
 			__be32 wrid_hi;
 			__be32 wrid_low;
 		} gen;
+		u64 drain_cookie;
 	} u;
 	__be64 reserved;
 	__be64 bits_type_ts;
@@ -238,6 +239,7 @@ struct t4_cqe {
 /* generic accessor macros */
 #define CQE_WRID_HI(x)		(be32_to_cpu((x)->u.gen.wrid_hi))
 #define CQE_WRID_LOW(x)		(be32_to_cpu((x)->u.gen.wrid_low))
+#define CQE_DRAIN_COOKIE(x)	((x)->u.drain_cookie)
 
 /* macros for flit 3 of the cqe */
 #define CQE_GENBIT_S	63