[PATCH] IB/mthca: Add SRQ implementation

Add mthca support for shared receive queues (SRQs),
including userspace SRQs.
Signed-off-by: Roland Dreier <rolandd@cisco.com>

drivers/infiniband/hw/mthca/Makefile

@@ -9,4 +9,4 @@ obj-$(CONFIG_INFINIBAND_MTHCA) += ib_mthca.o
 ib_mthca-y :=	mthca_main.o mthca_cmd.o mthca_profile.o mthca_reset.o \
 		mthca_allocator.o mthca_eq.o mthca_pd.o mthca_cq.o \
 		mthca_mr.o mthca_qp.o mthca_av.o mthca_mcg.o mthca_mad.o \
-		mthca_provider.o mthca_memfree.o mthca_uar.o
+		mthca_provider.o mthca_memfree.o mthca_uar.o mthca_srq.o

drivers/infiniband/hw/mthca/mthca_cmd.c

@@ -109,6 +109,7 @@ enum {
 	CMD_SW2HW_SRQ 	    = 0x35,
 	CMD_HW2SW_SRQ 	    = 0x36,
 	CMD_QUERY_SRQ       = 0x37,
+	CMD_ARM_SRQ         = 0x40,
 
 	/* QP/EE commands */
 	CMD_RST2INIT_QPEE   = 0x19,
@@ -1032,6 +1033,8 @@ int mthca_QUERY_DEV_LIM(struct mthca_dev *dev,
 
 	mthca_dbg(dev, "Max QPs: %d, reserved QPs: %d, entry size: %d\n",
 		  dev_lim->max_qps, dev_lim->reserved_qps, dev_lim->qpc_entry_sz);
+	mthca_dbg(dev, "Max SRQs: %d, reserved SRQs: %d, entry size: %d\n",
+		  dev_lim->max_srqs, dev_lim->reserved_srqs, dev_lim->srq_entry_sz);
 	mthca_dbg(dev, "Max CQs: %d, reserved CQs: %d, entry size: %d\n",
 		  dev_lim->max_cqs, dev_lim->reserved_cqs, dev_lim->cqc_entry_sz);
 	mthca_dbg(dev, "Max EQs: %d, reserved EQs: %d, entry size: %d\n",
@@ -1500,6 +1503,27 @@ int mthca_HW2SW_CQ(struct mthca_dev *dev, struct mthca_mailbox *mailbox,
 			     CMD_TIME_CLASS_A, status);
 }
 
+int mthca_SW2HW_SRQ(struct mthca_dev *dev, struct mthca_mailbox *mailbox,
+		    int srq_num, u8 *status)
+{
+	return mthca_cmd(dev, mailbox->dma, srq_num, 0, CMD_SW2HW_SRQ,
+			CMD_TIME_CLASS_A, status);
+}
+
+int mthca_HW2SW_SRQ(struct mthca_dev *dev, struct mthca_mailbox *mailbox,
+		    int srq_num, u8 *status)
+{
+	return mthca_cmd_box(dev, 0, mailbox->dma, srq_num, 0,
+			     CMD_HW2SW_SRQ,
+			     CMD_TIME_CLASS_A, status);
+}
+
+int mthca_ARM_SRQ(struct mthca_dev *dev, int srq_num, int limit, u8 *status)
+{
+	return mthca_cmd(dev, limit, srq_num, 0, CMD_ARM_SRQ,
+			 CMD_TIME_CLASS_B, status);
+}
+
 int mthca_MODIFY_QP(struct mthca_dev *dev, int trans, u32 num,
 		    int is_ee, struct mthca_mailbox *mailbox, u32 optmask,
 		    u8 *status)

drivers/infiniband/hw/mthca/mthca_cmd.h

@@ -298,6 +298,11 @@ int mthca_SW2HW_CQ(struct mthca_dev *dev, struct mthca_mailbox *mailbox,
 		   int cq_num, u8 *status);
 int mthca_HW2SW_CQ(struct mthca_dev *dev, struct mthca_mailbox *mailbox,
 		   int cq_num, u8 *status);
+int mthca_SW2HW_SRQ(struct mthca_dev *dev, struct mthca_mailbox *mailbox,
+		    int srq_num, u8 *status);
+int mthca_HW2SW_SRQ(struct mthca_dev *dev, struct mthca_mailbox *mailbox,
+		    int srq_num, u8 *status);
+int mthca_ARM_SRQ(struct mthca_dev *dev, int srq_num, int limit, u8 *status);
 int mthca_MODIFY_QP(struct mthca_dev *dev, int trans, u32 num,
 		    int is_ee, struct mthca_mailbox *mailbox, u32 optmask,
 		    u8 *status);

drivers/infiniband/hw/mthca/mthca_cq.c

@@ -224,7 +224,8 @@ void mthca_cq_event(struct mthca_dev *dev, u32 cqn)
 	cq->ibcq.comp_handler(&cq->ibcq, cq->ibcq.cq_context);
 }
 
-void mthca_cq_clean(struct mthca_dev *dev, u32 cqn, u32 qpn)
+void mthca_cq_clean(struct mthca_dev *dev, u32 cqn, u32 qpn,
+		    struct mthca_srq *srq)
 {
 	struct mthca_cq *cq;
 	struct mthca_cqe *cqe;
@@ -265,8 +266,11 @@ void mthca_cq_clean(struct mthca_dev *dev, u32 cqn, u32 qpn)
 	 */
 	while (prod_index > cq->cons_index) {
 		cqe = get_cqe(cq, (prod_index - 1) & cq->ibcq.cqe);
-		if (cqe->my_qpn == cpu_to_be32(qpn))
+		if (cqe->my_qpn == cpu_to_be32(qpn)) {
+			if (srq)
+				mthca_free_srq_wqe(srq, be32_to_cpu(cqe->wqe));
 			++nfreed;
+		}
 		else if (nfreed)
 			memcpy(get_cqe(cq, (prod_index - 1 + nfreed) &
 				       cq->ibcq.cqe),
@@ -455,23 +459,27 @@ static inline int mthca_poll_one(struct mthca_dev *dev,
 			     >> wq->wqe_shift);
 		entry->wr_id = (*cur_qp)->wrid[wqe_index +
 					       (*cur_qp)->rq.max];
+	} else if ((*cur_qp)->ibqp.srq) {
+		struct mthca_srq *srq = to_msrq((*cur_qp)->ibqp.srq);
+		u32 wqe = be32_to_cpu(cqe->wqe);
+		wq = NULL;
+		wqe_index = wqe >> srq->wqe_shift;
+		entry->wr_id = srq->wrid[wqe_index];
+		mthca_free_srq_wqe(srq, wqe);
 	} else {
 		wq = &(*cur_qp)->rq;
 		wqe_index = be32_to_cpu(cqe->wqe) >> wq->wqe_shift;
 		entry->wr_id = (*cur_qp)->wrid[wqe_index];
 	}
 
-	if (wq->last_comp < wqe_index)
-		wq->tail += wqe_index - wq->last_comp;
-	else
-		wq->tail += wqe_index + wq->max - wq->last_comp;
-
-	wq->last_comp = wqe_index;
+	if (wq) {
+		if (wq->last_comp < wqe_index)
+			wq->tail += wqe_index - wq->last_comp;
+		else
+			wq->tail += wqe_index + wq->max - wq->last_comp;
 
-	if (0)
-		mthca_dbg(dev, "%s completion for QP %06x, index %d (nr %d)\n",
-			  is_send ? "Send" : "Receive",
-			  (*cur_qp)->qpn, wqe_index, wq->max);
+		wq->last_comp = wqe_index;
+	}
 
 	if (is_error) {
 		err = handle_error_cqe(dev, cq, *cur_qp, wqe_index, is_send,

drivers/infiniband/hw/mthca/mthca_dev.h

@@ -218,6 +218,13 @@ struct mthca_cq_table {
 	struct mthca_icm_table *table;
 };
 
+struct mthca_srq_table {
+	struct mthca_alloc 	alloc;
+	spinlock_t         	lock;
+	struct mthca_array      srq;
+	struct mthca_icm_table *table;
+};
+
 struct mthca_qp_table {
 	struct mthca_alloc     	alloc;
 	u32                    	rdb_base;
@@ -299,6 +306,7 @@ struct mthca_dev {
 	struct mthca_mr_table  mr_table;
 	struct mthca_eq_table  eq_table;
 	struct mthca_cq_table  cq_table;
+	struct mthca_srq_table srq_table;
 	struct mthca_qp_table  qp_table;
 	struct mthca_av_table  av_table;
 	struct mthca_mcg_table mcg_table;
@@ -372,6 +380,7 @@ int mthca_init_pd_table(struct mthca_dev *dev);
 int mthca_init_mr_table(struct mthca_dev *dev);
 int mthca_init_eq_table(struct mthca_dev *dev);
 int mthca_init_cq_table(struct mthca_dev *dev);
+int mthca_init_srq_table(struct mthca_dev *dev);
 int mthca_init_qp_table(struct mthca_dev *dev);
 int mthca_init_av_table(struct mthca_dev *dev);
 int mthca_init_mcg_table(struct mthca_dev *dev);
@@ -381,6 +390,7 @@ void mthca_cleanup_pd_table(struct mthca_dev *dev);
 void mthca_cleanup_mr_table(struct mthca_dev *dev);
 void mthca_cleanup_eq_table(struct mthca_dev *dev);
 void mthca_cleanup_cq_table(struct mthca_dev *dev);
+void mthca_cleanup_srq_table(struct mthca_dev *dev);
 void mthca_cleanup_qp_table(struct mthca_dev *dev);
 void mthca_cleanup_av_table(struct mthca_dev *dev);
 void mthca_cleanup_mcg_table(struct mthca_dev *dev);
@@ -431,7 +441,19 @@ int mthca_init_cq(struct mthca_dev *dev, int nent,
 void mthca_free_cq(struct mthca_dev *dev,
 		   struct mthca_cq *cq);
 void mthca_cq_event(struct mthca_dev *dev, u32 cqn);
-void mthca_cq_clean(struct mthca_dev *dev, u32 cqn, u32 qpn);
+void mthca_cq_clean(struct mthca_dev *dev, u32 cqn, u32 qpn,
+		    struct mthca_srq *srq);
+
+int mthca_alloc_srq(struct mthca_dev *dev, struct mthca_pd *pd,
+		    struct ib_srq_attr *attr, struct mthca_srq *srq);
+void mthca_free_srq(struct mthca_dev *dev, struct mthca_srq *srq);
+void mthca_srq_event(struct mthca_dev *dev, u32 srqn,
+		     enum ib_event_type event_type);
+void mthca_free_srq_wqe(struct mthca_srq *srq, u32 wqe_addr);
+int mthca_tavor_post_srq_recv(struct ib_srq *srq, struct ib_recv_wr *wr,
+			      struct ib_recv_wr **bad_wr);
+int mthca_arbel_post_srq_recv(struct ib_srq *srq, struct ib_recv_wr *wr,
+			      struct ib_recv_wr **bad_wr);
 
 void mthca_qp_event(struct mthca_dev *dev, u32 qpn,
 		    enum ib_event_type event_type);

drivers/infiniband/hw/mthca/mthca_main.c

@@ -253,6 +253,8 @@ static int __devinit mthca_init_tavor(struct mthca_dev *mdev)
 	profile = default_profile;
 	profile.num_uar   = dev_lim.uar_size / PAGE_SIZE;
 	profile.uarc_size = 0;
+	if (mdev->mthca_flags & MTHCA_FLAG_SRQ)
+		profile.num_srq = dev_lim.max_srqs;
 
 	err = mthca_make_profile(mdev, &profile, &dev_lim, &init_hca);
 	if (err < 0)
@@ -424,15 +426,29 @@ static int __devinit mthca_init_icm(struct mthca_dev *mdev,
 	}
 
        mdev->cq_table.table = mthca_alloc_icm_table(mdev, init_hca->cqc_base,
-						     dev_lim->cqc_entry_sz,
-						     mdev->limits.num_cqs,
-						     mdev->limits.reserved_cqs, 0);
+						    dev_lim->cqc_entry_sz,
+						    mdev->limits.num_cqs,
+						    mdev->limits.reserved_cqs, 0);
 	if (!mdev->cq_table.table) {
 		mthca_err(mdev, "Failed to map CQ context memory, aborting.\n");
 		err = -ENOMEM;
 		goto err_unmap_rdb;
 	}
 
+	if (mdev->mthca_flags & MTHCA_FLAG_SRQ) {
+		mdev->srq_table.table =
+			mthca_alloc_icm_table(mdev, init_hca->srqc_base,
+					      dev_lim->srq_entry_sz,
+					      mdev->limits.num_srqs,
+					      mdev->limits.reserved_srqs, 0);
+		if (!mdev->srq_table.table) {
+			mthca_err(mdev, "Failed to map SRQ context memory, "
+				  "aborting.\n");
+			err = -ENOMEM;
+			goto err_unmap_cq;
+		}
+	}
+
 	/*
 	 * It's not strictly required, but for simplicity just map the
 	 * whole multicast group table now.  The table isn't very big
@@ -448,11 +464,15 @@ static int __devinit mthca_init_icm(struct mthca_dev *mdev,
 	if (!mdev->mcg_table.table) {
 		mthca_err(mdev, "Failed to map MCG context memory, aborting.\n");
 		err = -ENOMEM;
-		goto err_unmap_cq;
+		goto err_unmap_srq;
 	}
 
 	return 0;
 
+err_unmap_srq:
+	if (mdev->mthca_flags & MTHCA_FLAG_SRQ)
+		mthca_free_icm_table(mdev, mdev->srq_table.table);
+
 err_unmap_cq:
 	mthca_free_icm_table(mdev, mdev->cq_table.table);
 
@@ -532,6 +552,8 @@ static int __devinit mthca_init_arbel(struct mthca_dev *mdev)
 	profile = default_profile;
 	profile.num_uar  = dev_lim.uar_size / PAGE_SIZE;
 	profile.num_udav = 0;
+	if (mdev->mthca_flags & MTHCA_FLAG_SRQ)
+		profile.num_srq = dev_lim.max_srqs;
 
 	icm_size = mthca_make_profile(mdev, &profile, &dev_lim, &init_hca);
 	if ((int) icm_size < 0) {
@@ -558,6 +580,8 @@ static int __devinit mthca_init_arbel(struct mthca_dev *mdev)
 	return 0;
 
 err_free_icm:
+	if (mdev->mthca_flags & MTHCA_FLAG_SRQ)
+		mthca_free_icm_table(mdev, mdev->srq_table.table);
 	mthca_free_icm_table(mdev, mdev->cq_table.table);
 	mthca_free_icm_table(mdev, mdev->qp_table.rdb_table);
 	mthca_free_icm_table(mdev, mdev->qp_table.eqp_table);
@@ -587,6 +611,8 @@ static void mthca_close_hca(struct mthca_dev *mdev)
 	mthca_CLOSE_HCA(mdev, 0, &status);
 
 	if (mthca_is_memfree(mdev)) {
+		if (mdev->mthca_flags & MTHCA_FLAG_SRQ)
+			mthca_free_icm_table(mdev, mdev->srq_table.table);
 		mthca_free_icm_table(mdev, mdev->cq_table.table);
 		mthca_free_icm_table(mdev, mdev->qp_table.rdb_table);
 		mthca_free_icm_table(mdev, mdev->qp_table.eqp_table);
@@ -731,11 +757,18 @@ static int __devinit mthca_setup_hca(struct mthca_dev *dev)
 		goto err_cmd_poll;
 	}
 
+	err = mthca_init_srq_table(dev);
+	if (err) {
+		mthca_err(dev, "Failed to initialize "
+			  "shared receive queue table, aborting.\n");
+		goto err_cq_table_free;
+	}
+
 	err = mthca_init_qp_table(dev);
 	if (err) {
 		mthca_err(dev, "Failed to initialize "
 			  "queue pair table, aborting.\n");
-		goto err_cq_table_free;
+		goto err_srq_table_free;
 	}
 
 	err = mthca_init_av_table(dev);
@@ -760,6 +793,9 @@ static int __devinit mthca_setup_hca(struct mthca_dev *dev)
 err_qp_table_free:
 	mthca_cleanup_qp_table(dev);
 
+err_srq_table_free:
+	mthca_cleanup_srq_table(dev);
+
 err_cq_table_free:
 	mthca_cleanup_cq_table(dev);
 
@@ -1046,6 +1082,7 @@ static int __devinit mthca_init_one(struct pci_dev *pdev,
 	mthca_cleanup_mcg_table(mdev);
 	mthca_cleanup_av_table(mdev);
 	mthca_cleanup_qp_table(mdev);
+	mthca_cleanup_srq_table(mdev);
 	mthca_cleanup_cq_table(mdev);
 	mthca_cmd_use_polling(mdev);
 	mthca_cleanup_eq_table(mdev);
@@ -1095,6 +1132,7 @@ static void __devexit mthca_remove_one(struct pci_dev *pdev)
 		mthca_cleanup_mcg_table(mdev);
 		mthca_cleanup_av_table(mdev);
 		mthca_cleanup_qp_table(mdev);
+		mthca_cleanup_srq_table(mdev);
 		mthca_cleanup_cq_table(mdev);
 		mthca_cmd_use_polling(mdev);
 		mthca_cleanup_eq_table(mdev);

drivers/infiniband/hw/mthca/mthca_profile.c

@@ -102,6 +102,7 @@ u64 mthca_make_profile(struct mthca_dev *dev,
 	profile[MTHCA_RES_UARC].size = request->uarc_size;
 
 	profile[MTHCA_RES_QP].num    = request->num_qp;
+	profile[MTHCA_RES_SRQ].num   = request->num_srq;
 	profile[MTHCA_RES_EQP].num   = request->num_qp;
 	profile[MTHCA_RES_RDB].num   = request->num_qp * request->rdb_per_qp;
 	profile[MTHCA_RES_CQ].num    = request->num_cq;

drivers/infiniband/hw/mthca/mthca_profile.h

@@ -42,6 +42,7 @@
 struct mthca_profile {
 	int num_qp;
 	int rdb_per_qp;
+	int num_srq;
 	int num_cq;
 	int num_mcg;
 	int num_mpt;

drivers/infiniband/hw/mthca/mthca_provider.c

@@ -425,6 +425,77 @@ static int mthca_ah_destroy(struct ib_ah *ah)
 	return 0;
 }
 
+static struct ib_srq *mthca_create_srq(struct ib_pd *pd,
+				       struct ib_srq_init_attr *init_attr,
+				       struct ib_udata *udata)
+{
+	struct mthca_create_srq ucmd;
+	struct mthca_ucontext *context = NULL;
+	struct mthca_srq *srq;
+	int err;
+
+	srq = kmalloc(sizeof *srq, GFP_KERNEL);
+	if (!srq)
+		return ERR_PTR(-ENOMEM);
+
+	if (pd->uobject) {
+		context = to_mucontext(pd->uobject->context);
+
+		if (ib_copy_from_udata(&ucmd, udata, sizeof ucmd))
+			return ERR_PTR(-EFAULT);
+
+		err = mthca_map_user_db(to_mdev(pd->device), &context->uar,
+					context->db_tab, ucmd.db_index,
+					ucmd.db_page);
+
+		if (err)
+			goto err_free;
+
+		srq->mr.ibmr.lkey = ucmd.lkey;
+		srq->db_index     = ucmd.db_index;
+	}
+
+	err = mthca_alloc_srq(to_mdev(pd->device), to_mpd(pd),
+			      &init_attr->attr, srq);
+
+	if (err && pd->uobject)
+		mthca_unmap_user_db(to_mdev(pd->device), &context->uar,
+				    context->db_tab, ucmd.db_index);
+
+	if (err)
+		goto err_free;
+
+	if (context && ib_copy_to_udata(udata, &srq->srqn, sizeof (__u32))) {
+		mthca_free_srq(to_mdev(pd->device), srq);
+		err = -EFAULT;
+		goto err_free;
+	}
+
+	return &srq->ibsrq;
+
+err_free:
+	kfree(srq);
+
+	return ERR_PTR(err);
+}
+
+static int mthca_destroy_srq(struct ib_srq *srq)
+{
+	struct mthca_ucontext *context;
+
+	if (srq->uobject) {
+		context = to_mucontext(srq->uobject->context);
+
+		mthca_unmap_user_db(to_mdev(srq->device), &context->uar,
+				    context->db_tab, to_msrq(srq)->db_index);
+	}
+
+	mthca_free_srq(to_mdev(srq->device), to_msrq(srq));
+	kfree(srq);
+
+	return 0;
+}
+
 static struct ib_qp *mthca_create_qp(struct ib_pd *pd,
 				     struct ib_qp_init_attr *init_attr,
 				     struct ib_udata *udata)
@@ -1003,6 +1074,17 @@ int mthca_register_device(struct mthca_dev *dev)
 	dev->ib_dev.dealloc_pd           = mthca_dealloc_pd;
 	dev->ib_dev.create_ah            = mthca_ah_create;
 	dev->ib_dev.destroy_ah           = mthca_ah_destroy;
+
+	if (dev->mthca_flags & MTHCA_FLAG_SRQ) {
+		dev->ib_dev.create_srq           = mthca_create_srq;
+		dev->ib_dev.destroy_srq          = mthca_destroy_srq;
+
+		if (mthca_is_memfree(dev))
+			dev->ib_dev.post_srq_recv = mthca_arbel_post_srq_recv;
+		else
+			dev->ib_dev.post_srq_recv = mthca_tavor_post_srq_recv;
+	}
+
 	dev->ib_dev.create_qp            = mthca_create_qp;
 	dev->ib_dev.modify_qp            = mthca_modify_qp;
 	dev->ib_dev.destroy_qp           = mthca_destroy_qp;

drivers/infiniband/hw/mthca/mthca_provider.h

@@ -197,6 +197,29 @@ struct mthca_cq {
 	wait_queue_head_t      wait;
 };
 
+struct mthca_srq {
+	struct ib_srq		ibsrq;
+	spinlock_t		lock;
+	atomic_t		refcount;
+	int			srqn;
+	int			max;
+	int			max_gs;
+	int			wqe_shift;
+	int			first_free;
+	int			last_free;
+	u16			counter;  /* Arbel only */
+	int			db_index; /* Arbel only */
+	__be32		       *db;       /* Arbel only */
+	void		       *last;
+
+	int			is_direct;
+	u64		       *wrid;
+	union mthca_buf		queue;
+	struct mthca_mr		mr;
+
+	wait_queue_head_t	wait;
+};
+
 struct mthca_wq {
 	spinlock_t lock;
 	int        max;
@@ -277,6 +300,11 @@ static inline struct mthca_cq *to_mcq(struct ib_cq *ibcq)
 	return container_of(ibcq, struct mthca_cq, ibcq);
 }
 
+static inline struct mthca_srq *to_msrq(struct ib_srq *ibsrq)
+{
+	return container_of(ibsrq, struct mthca_srq, ibsrq);
+}
+
 static inline struct mthca_qp *to_mqp(struct ib_qp *ibqp)
 {
 	return container_of(ibqp, struct mthca_qp, ibqp);

drivers/infiniband/hw/mthca/mthca_qp.c

@@ -612,10 +612,13 @@ int mthca_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr, int attr_mask)
 		qp_context->mtu_msgmax = (attr->path_mtu << 5) | 31;
 
 	if (mthca_is_memfree(dev)) {
-		qp_context->rq_size_stride =
-			((ffs(qp->rq.max) - 1) << 3) | (qp->rq.wqe_shift - 4);
-		qp_context->sq_size_stride =
-			((ffs(qp->sq.max) - 1) << 3) | (qp->sq.wqe_shift - 4);
+		if (qp->rq.max)
+			qp_context->rq_size_stride = long_log2(qp->rq.max) << 3;
+		qp_context->rq_size_stride |= qp->rq.wqe_shift - 4;
+
+		if (qp->sq.max)
+			qp_context->sq_size_stride = long_log2(qp->sq.max) << 3;
+		qp_context->sq_size_stride |= qp->sq.wqe_shift - 4;
 	}
 
 	/* leave arbel_sched_queue as 0 */
@@ -784,6 +787,9 @@ int mthca_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr, int attr_mask)
 
 	qp_context->params2 |= cpu_to_be32(MTHCA_QP_BIT_RSC);
 
+	if (ibqp->srq)
+		qp_context->params2 |= cpu_to_be32(MTHCA_QP_BIT_RIC);
+
 	if (attr_mask & IB_QP_MIN_RNR_TIMER) {
 		qp_context->rnr_nextrecvpsn |= cpu_to_be32(attr->min_rnr_timer << 24);
 		qp_param->opt_param_mask |= cpu_to_be32(MTHCA_QP_OPTPAR_RNR_TIMEOUT);
@@ -806,6 +812,10 @@ int mthca_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr, int attr_mask)
 		qp_param->opt_param_mask |= cpu_to_be32(MTHCA_QP_OPTPAR_Q_KEY);
 	}
 
+	if (ibqp->srq)
+		qp_context->srqn = cpu_to_be32(1 << 24 |
+					       to_msrq(ibqp->srq)->srqn);
+
 	err = mthca_MODIFY_QP(dev, state_table[cur_state][new_state].trans,
 			      qp->qpn, 0, mailbox, 0, &status);
 	if (status) {
@@ -1260,9 +1270,11 @@ void mthca_free_qp(struct mthca_dev *dev,
 	 * unref the mem-free tables and free the QPN in our table.
 	 */
 	if (!qp->ibqp.uobject) {
-		mthca_cq_clean(dev, to_mcq(qp->ibqp.send_cq)->cqn, qp->qpn);
+		mthca_cq_clean(dev, to_mcq(qp->ibqp.send_cq)->cqn, qp->qpn,
+			       qp->ibqp.srq ? to_msrq(qp->ibqp.srq) : NULL);
 		if (qp->ibqp.send_cq != qp->ibqp.recv_cq)
-			mthca_cq_clean(dev, to_mcq(qp->ibqp.recv_cq)->cqn, qp->qpn);
+			mthca_cq_clean(dev, to_mcq(qp->ibqp.recv_cq)->cqn, qp->qpn,
+				       qp->ibqp.srq ? to_msrq(qp->ibqp.srq) : NULL);
 
 		mthca_free_memfree(dev, qp);
 		mthca_free_wqe_buf(dev, qp);
@@ -2008,6 +2020,15 @@ int mthca_free_err_wqe(struct mthca_dev *dev, struct mthca_qp *qp, int is_send,
 {
 	struct mthca_next_seg *next;
 
+	/*
+	 * For SRQs, all WQEs generate a CQE, so we're always at the
+	 * end of the doorbell chain.
+	 */
+	if (qp->ibqp.srq) {
+		*new_wqe = 0;
+		return 0;
+	}
+
 	if (is_send)
 		next = get_send_wqe(qp, index);
 	else

drivers/infiniband/hw/mthca/mthca_user.h

@@ -69,6 +69,17 @@ struct mthca_create_cq_resp {
 	__u32 reserved;
 };
 
+struct mthca_create_srq {
+	__u32 lkey;
+	__u32 db_index;
+	__u64 db_page;
+};
+
+struct mthca_create_srq_resp {
+	__u32 srqn;
+	__u32 reserved;
+};
+
 struct mthca_create_qp {
 	__u32 lkey;
 	__u32 reserved;