IB/core: Add "type 2" memory windows support

This patch enhances the IB core support for Memory Windows (MWs).

MWs allow an application to have better/flexible control over remote
access to memory.

Two types of MWs are supported, with the second type having two flavors:

    Type 1  - associated with PD only
    Type 2A - associated with QPN only
    Type 2B - associated with PD and QPN

Applications can allocate a MW once, and then repeatedly bind the MW
to different ranges in MRs that are associated to the same PD. Type 1
windows are bound through a verb, while type 2 windows are bound by
posting a work request.

The 32-bit memory key is composed of a 24-bit index and an 8-bit
key. The key is changed with each bind, thus allowing more control
over the peer's use of the memory key.

The changes introduced are the following:

* add memory window type enum and a corresponding parameter to ib_alloc_mw.
* type 2 memory window bind work request support.
* create a struct that contains the common part of the bind verb struct
  ibv_mw_bind and the bind work request into a single struct.
* add the ib_inc_rkey helper function to advance the tag part of an rkey.

Consumer interface details:

* new device capability flags IB_DEVICE_MEM_WINDOW_TYPE_2A and
  IB_DEVICE_MEM_WINDOW_TYPE_2B are added to indicate device support
  for these features.

  Devices can set either IB_DEVICE_MEM_WINDOW_TYPE_2A or
  IB_DEVICE_MEM_WINDOW_TYPE_2B if it supports type 2A or type 2B
  memory windows. It can set neither to indicate it doesn't support
  type 2 windows at all.

* modify existing provides and consumers code to the new param of
  ib_alloc_mw and the ib_mw_bind_info structure

Signed-off-by: Haggai Eran <haggaie@mellanox.com>
Signed-off-by: Shani Michaeli <shanim@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>

10 files changed, 110 insertions, 41 deletions

diff --git a/drivers/infiniband/core/verbs.c b/drivers/infiniband/core/verbs.c
index 30f199e8579f..a8fdd3381405 100644
--- a/drivers/infiniband/core/verbs.c
+++ b/drivers/infiniband/core/verbs.c
@@ -1099,18 +1099,19 @@ EXPORT_SYMBOL(ib_free_fast_reg_page_list);
 
 /* Memory windows */
 
-struct ib_mw *ib_alloc_mw(struct ib_pd *pd)
+struct ib_mw *ib_alloc_mw(struct ib_pd *pd, enum ib_mw_type type)
 {
 	struct ib_mw *mw;
 
 	if (!pd->device->alloc_mw)
 		return ERR_PTR(-ENOSYS);
 
-	mw = pd->device->alloc_mw(pd);
+	mw = pd->device->alloc_mw(pd, type);
 	if (!IS_ERR(mw)) {
 		mw->device  = pd->device;
 		mw->pd      = pd;
 		mw->uobject = NULL;
+		mw->type    = type;
 		atomic_inc(&pd->usecnt);
 	}
 
diff --git a/drivers/infiniband/hw/cxgb3/iwch_provider.c b/drivers/infiniband/hw/cxgb3/iwch_provider.c
index 0bdf09aa6f42..074d5c25f50b 100644
--- a/drivers/infiniband/hw/cxgb3/iwch_provider.c
+++ b/drivers/infiniband/hw/cxgb3/iwch_provider.c
@@ -738,7 +738,7 @@ static struct ib_mr *iwch_get_dma_mr(struct ib_pd *pd, int acc)
 	return ibmr;
 }
 
-static struct ib_mw *iwch_alloc_mw(struct ib_pd *pd)
+static struct ib_mw *iwch_alloc_mw(struct ib_pd *pd, enum ib_mw_type type)
 {
 	struct iwch_dev *rhp;
 	struct iwch_pd *php;
@@ -747,6 +747,9 @@ static struct ib_mw *iwch_alloc_mw(struct ib_pd *pd)
 	u32 stag = 0;
 	int ret;
 
+	if (type != IB_MW_TYPE_1)
+		return ERR_PTR(-EINVAL);
+
 	php = to_iwch_pd(pd);
 	rhp = php->rhp;
 	mhp = kzalloc(sizeof(*mhp), GFP_KERNEL);
diff --git a/drivers/infiniband/hw/cxgb3/iwch_qp.c b/drivers/infiniband/hw/cxgb3/iwch_qp.c
index 6de8463f453b..e5649e8b215d 100644
--- a/drivers/infiniband/hw/cxgb3/iwch_qp.c
+++ b/drivers/infiniband/hw/cxgb3/iwch_qp.c
@@ -567,18 +567,19 @@ int iwch_bind_mw(struct ib_qp *qp,
 	if (mw_bind->send_flags & IB_SEND_SIGNALED)
 		t3_wr_flags = T3_COMPLETION_FLAG;
 
-	sgl.addr = mw_bind->addr;
-	sgl.lkey = mw_bind->mr->lkey;
-	sgl.length = mw_bind->length;
+	sgl.addr = mw_bind->bind_info.addr;
+	sgl.lkey = mw_bind->bind_info.mr->lkey;
+	sgl.length = mw_bind->bind_info.length;
 	wqe->bind.reserved = 0;
 	wqe->bind.type = TPT_VATO;
 
 	/* TBD: check perms */
-	wqe->bind.perms = iwch_ib_to_tpt_bind_access(mw_bind->mw_access_flags);
-	wqe->bind.mr_stag = cpu_to_be32(mw_bind->mr->lkey);
+	wqe->bind.perms = iwch_ib_to_tpt_bind_access(
+		mw_bind->bind_info.mw_access_flags);
+	wqe->bind.mr_stag = cpu_to_be32(mw_bind->bind_info.mr->lkey);
 	wqe->bind.mw_stag = cpu_to_be32(mw->rkey);
-	wqe->bind.mw_len = cpu_to_be32(mw_bind->length);
-	wqe->bind.mw_va = cpu_to_be64(mw_bind->addr);
+	wqe->bind.mw_len = cpu_to_be32(mw_bind->bind_info.length);
+	wqe->bind.mw_va = cpu_to_be64(mw_bind->bind_info.addr);
 	err = iwch_sgl2pbl_map(rhp, &sgl, 1, &pbl_addr, &page_size);
 	if (err) {
 		spin_unlock_irqrestore(&qhp->lock, flag);
diff --git a/drivers/infiniband/hw/cxgb4/iw_cxgb4.h b/drivers/infiniband/hw/cxgb4/iw_cxgb4.h
index 9c1644fb0259..c8bd239915d5 100644
--- a/drivers/infiniband/hw/cxgb4/iw_cxgb4.h
+++ b/drivers/infiniband/hw/cxgb4/iw_cxgb4.h
@@ -866,7 +866,7 @@ struct ib_fast_reg_page_list *c4iw_alloc_fastreg_pbl(
 					int page_list_len);
 struct ib_mr *c4iw_alloc_fast_reg_mr(struct ib_pd *pd, int pbl_depth);
 int c4iw_dealloc_mw(struct ib_mw *mw);
-struct ib_mw *c4iw_alloc_mw(struct ib_pd *pd);
+struct ib_mw *c4iw_alloc_mw(struct ib_pd *pd, enum ib_mw_type type);
 struct ib_mr *c4iw_reg_user_mr(struct ib_pd *pd, u64 start,
 					   u64 length, u64 virt, int acc,
 					   struct ib_udata *udata);
diff --git a/drivers/infiniband/hw/cxgb4/mem.c b/drivers/infiniband/hw/cxgb4/mem.c
index afd81790ab3c..903a92d6f91d 100644
--- a/drivers/infiniband/hw/cxgb4/mem.c
+++ b/drivers/infiniband/hw/cxgb4/mem.c
@@ -650,7 +650,7 @@ err:
 	return ERR_PTR(err);
 }
 
-struct ib_mw *c4iw_alloc_mw(struct ib_pd *pd)
+struct ib_mw *c4iw_alloc_mw(struct ib_pd *pd, enum ib_mw_type type)
 {
 	struct c4iw_dev *rhp;
 	struct c4iw_pd *php;
@@ -659,6 +659,9 @@ struct ib_mw *c4iw_alloc_mw(struct ib_pd *pd)
 	u32 stag = 0;
 	int ret;
 
+	if (type != IB_MW_TYPE_1)
+		return ERR_PTR(-EINVAL);
+
 	php = to_c4iw_pd(pd);
 	rhp = php->rhp;
 	mhp = kzalloc(sizeof(*mhp), GFP_KERNEL);
diff --git a/drivers/infiniband/hw/ehca/ehca_iverbs.h b/drivers/infiniband/hw/ehca/ehca_iverbs.h
index 8f7f282ead65..22f79afa7fc1 100644
--- a/drivers/infiniband/hw/ehca/ehca_iverbs.h
+++ b/drivers/infiniband/hw/ehca/ehca_iverbs.h
@@ -95,7 +95,7 @@ int ehca_query_mr(struct ib_mr *mr, struct ib_mr_attr *mr_attr);
 
 int ehca_dereg_mr(struct ib_mr *mr);
 
-struct ib_mw *ehca_alloc_mw(struct ib_pd *pd);
+struct ib_mw *ehca_alloc_mw(struct ib_pd *pd, enum ib_mw_type type);
 
 int ehca_bind_mw(struct ib_qp *qp, struct ib_mw *mw,
 		 struct ib_mw_bind *mw_bind);
diff --git a/drivers/infiniband/hw/ehca/ehca_mrmw.c b/drivers/infiniband/hw/ehca/ehca_mrmw.c
index 87844869dcc2..bcfb0c183620 100644
--- a/drivers/infiniband/hw/ehca/ehca_mrmw.c
+++ b/drivers/infiniband/hw/ehca/ehca_mrmw.c
@@ -688,7 +688,7 @@ dereg_mr_exit0:
 
 /*----------------------------------------------------------------------*/
 
-struct ib_mw *ehca_alloc_mw(struct ib_pd *pd)
+struct ib_mw *ehca_alloc_mw(struct ib_pd *pd, enum ib_mw_type type)
 {
 	struct ib_mw *ib_mw;
 	u64 h_ret;
@@ -698,6 +698,9 @@ struct ib_mw *ehca_alloc_mw(struct ib_pd *pd)
 		container_of(pd->device, struct ehca_shca, ib_device);
 	struct ehca_mw_hipzout_parms hipzout;
 
+	if (type != IB_MW_TYPE_1)
+		return ERR_PTR(-EINVAL);
+
 	e_mw = ehca_mw_new();
 	if (!e_mw) {
 		ib_mw = ERR_PTR(-ENOMEM);
diff --git a/drivers/infiniband/hw/nes/nes_verbs.c b/drivers/infiniband/hw/nes/nes_verbs.c
index 07e4fbad987a..8f67fe2e91e6 100644
--- a/drivers/infiniband/hw/nes/nes_verbs.c
+++ b/drivers/infiniband/hw/nes/nes_verbs.c
@@ -55,7 +55,8 @@ static void nes_unregister_ofa_device(struct nes_ib_device *nesibdev);
 /**
  * nes_alloc_mw
  */
-static struct ib_mw *nes_alloc_mw(struct ib_pd *ibpd) {
+static struct ib_mw *nes_alloc_mw(struct ib_pd *ibpd, enum ib_mw_type type)
+{
 	struct nes_pd *nespd = to_nespd(ibpd);
 	struct nes_vnic *nesvnic = to_nesvnic(ibpd->device);
 	struct nes_device *nesdev = nesvnic->nesdev;
@@ -71,6 +72,9 @@ static struct ib_mw *nes_alloc_mw(struct ib_pd *ibpd) {
 	u32 driver_key = 0;
 	u8 stag_key = 0;
 
+	if (type != IB_MW_TYPE_1)
+		return ERR_PTR(-EINVAL);
+
 	get_random_bytes(&next_stag_index, sizeof(next_stag_index));
 	stag_key = (u8)next_stag_index;
 
@@ -244,20 +248,19 @@ static int nes_bind_mw(struct ib_qp *ibqp, struct ib_mw *ibmw,
 	if (ibmw_bind->send_flags & IB_SEND_SIGNALED)
 		wqe_misc |= NES_IWARP_SQ_WQE_SIGNALED_COMPL;
 
-	if (ibmw_bind->mw_access_flags & IB_ACCESS_REMOTE_WRITE) {
+	if (ibmw_bind->bind_info.mw_access_flags & IB_ACCESS_REMOTE_WRITE)
 		wqe_misc |= NES_CQP_STAG_RIGHTS_REMOTE_WRITE;
-	}
-	if (ibmw_bind->mw_access_flags & IB_ACCESS_REMOTE_READ) {
+	if (ibmw_bind->bind_info.mw_access_flags & IB_ACCESS_REMOTE_READ)
 		wqe_misc |= NES_CQP_STAG_RIGHTS_REMOTE_READ;
-	}
 
 	set_wqe_32bit_value(wqe->wqe_words, NES_IWARP_SQ_WQE_MISC_IDX, wqe_misc);
-	set_wqe_32bit_value(wqe->wqe_words, NES_IWARP_SQ_BIND_WQE_MR_IDX, ibmw_bind->mr->lkey);
+	set_wqe_32bit_value(wqe->wqe_words, NES_IWARP_SQ_BIND_WQE_MR_IDX,
+			    ibmw_bind->bind_info.mr->lkey);
 	set_wqe_32bit_value(wqe->wqe_words, NES_IWARP_SQ_BIND_WQE_MW_IDX, ibmw->rkey);
 	set_wqe_32bit_value(wqe->wqe_words, NES_IWARP_SQ_BIND_WQE_LENGTH_LOW_IDX,
-			ibmw_bind->length);
+			ibmw_bind->bind_info.length);
 	wqe->wqe_words[NES_IWARP_SQ_BIND_WQE_LENGTH_HIGH_IDX] = 0;
-	u64temp = (u64)ibmw_bind->addr;
+	u64temp = (u64)ibmw_bind->bind_info.addr;
 	set_wqe_64bit_value(wqe->wqe_words, NES_IWARP_SQ_BIND_WQE_VA_FBO_LOW_IDX, u64temp);
 
 	head++;
diff --git a/include/rdma/ib_verbs.h b/include/rdma/ib_verbs.h
index 46bc045bbe15..98cc4b29fc5b 100644
--- a/include/rdma/ib_verbs.h
+++ b/include/rdma/ib_verbs.h
@@ -115,6 +115,8 @@ enum ib_device_cap_flags {
 	IB_DEVICE_XRC			= (1<<20),
 	IB_DEVICE_MEM_MGT_EXTENSIONS	= (1<<21),
 	IB_DEVICE_BLOCK_MULTICAST_LOOPBACK = (1<<22),
+	IB_DEVICE_MEM_WINDOW_TYPE_2A	= (1<<23),
+	IB_DEVICE_MEM_WINDOW_TYPE_2B	= (1<<24)
 };
 
 enum ib_atomic_cap {
@@ -715,6 +717,11 @@ enum ib_mig_state {
 	IB_MIG_ARMED
 };
 
+enum ib_mw_type {
+	IB_MW_TYPE_1 = 1,
+	IB_MW_TYPE_2 = 2
+};
+
 struct ib_qp_attr {
 	enum ib_qp_state	qp_state;
 	enum ib_qp_state	cur_qp_state;
@@ -758,6 +765,7 @@ enum ib_wr_opcode {
 	IB_WR_FAST_REG_MR,
 	IB_WR_MASKED_ATOMIC_CMP_AND_SWP,
 	IB_WR_MASKED_ATOMIC_FETCH_AND_ADD,
+	IB_WR_BIND_MW,
 };
 
 enum ib_send_flags {
@@ -780,6 +788,23 @@ struct ib_fast_reg_page_list {
 	unsigned int		max_page_list_len;
 };
 
+/**
+ * struct ib_mw_bind_info - Parameters for a memory window bind operation.
+ * @mr: A memory region to bind the memory window to.
+ * @addr: The address where the memory window should begin.
+ * @length: The length of the memory window, in bytes.
+ * @mw_access_flags: Access flags from enum ib_access_flags for the window.
+ *
+ * This struct contains the shared parameters for type 1 and type 2
+ * memory window bind operations.
+ */
+struct ib_mw_bind_info {
+	struct ib_mr   *mr;
+	u64		addr;
+	u64		length;
+	int		mw_access_flags;
+};
+
 struct ib_send_wr {
 	struct ib_send_wr      *next;
 	u64			wr_id;
@@ -823,6 +848,12 @@ struct ib_send_wr {
 			int				access_flags;
 			u32				rkey;
 		} fast_reg;
+		struct {
+			struct ib_mw            *mw;
+			/* The new rkey for the memory window. */
+			u32                      rkey;
+			struct ib_mw_bind_info   bind_info;
+		} bind_mw;
 	} wr;
 	u32			xrc_remote_srq_num;	/* XRC TGT QPs only */
 };
@@ -839,7 +870,8 @@ enum ib_access_flags {
 	IB_ACCESS_REMOTE_WRITE	= (1<<1),
 	IB_ACCESS_REMOTE_READ	= (1<<2),
 	IB_ACCESS_REMOTE_ATOMIC	= (1<<3),
-	IB_ACCESS_MW_BIND	= (1<<4)
+	IB_ACCESS_MW_BIND	= (1<<4),
+	IB_ZERO_BASED		= (1<<5)
 };
 
 struct ib_phys_buf {
@@ -862,13 +894,16 @@ enum ib_mr_rereg_flags {
 	IB_MR_REREG_ACCESS	= (1<<2)
 };
 
+/**
+ * struct ib_mw_bind - Parameters for a type 1 memory window bind operation.
+ * @wr_id:      Work request id.
+ * @send_flags: Flags from ib_send_flags enum.
+ * @bind_info:  More parameters of the bind operation.
+ */
 struct ib_mw_bind {
-	struct ib_mr   *mr;
-	u64		wr_id;
-	u64		addr;
-	u32		length;
-	int		send_flags;
-	int		mw_access_flags;
+	u64                    wr_id;
+	int                    send_flags;
+	struct ib_mw_bind_info bind_info;
 };
 
 struct ib_fmr_attr {
@@ -991,6 +1026,7 @@ struct ib_mw {
 	struct ib_pd		*pd;
 	struct ib_uobject	*uobject;
 	u32			rkey;
+	enum ib_mw_type         type;
 };
 
 struct ib_fmr {
@@ -1202,7 +1238,8 @@ struct ib_device {
 						    int num_phys_buf,
 						    int mr_access_flags,
 						    u64 *iova_start);
-	struct ib_mw *             (*alloc_mw)(struct ib_pd *pd);
+	struct ib_mw *             (*alloc_mw)(struct ib_pd *pd,
+					       enum ib_mw_type type);
 	int                        (*bind_mw)(struct ib_qp *qp,
 					      struct ib_mw *mw,
 					      struct ib_mw_bind *mw_bind);
@@ -2019,6 +2056,8 @@ int ib_query_mr(struct ib_mr *mr, struct ib_mr_attr *mr_attr);
  * ib_dereg_mr - Deregisters a memory region and removes it from the
  *   HCA translation table.
  * @mr: The memory region to deregister.
+ *
+ * This function can fail, if the memory region has memory windows bound to it.
  */
 int ib_dereg_mr(struct ib_mr *mr);
 
@@ -2071,10 +2110,22 @@ static inline void ib_update_fast_reg_key(struct ib_mr *mr, u8 newkey)
 }
 
 /**
+ * ib_inc_rkey - increments the key portion of the given rkey. Can be used
+ * for calculating a new rkey for type 2 memory windows.
+ * @rkey - the rkey to increment.
+ */
+static inline u32 ib_inc_rkey(u32 rkey)
+{
+	const u32 mask = 0x000000ff;
+	return ((rkey + 1) & mask) | (rkey & ~mask);
+}
+
+/**
  * ib_alloc_mw - Allocates a memory window.
  * @pd: The protection domain associated with the memory window.
+ * @type: The type of the memory window (1 or 2).
  */
-struct ib_mw *ib_alloc_mw(struct ib_pd *pd);
+struct ib_mw *ib_alloc_mw(struct ib_pd *pd, enum ib_mw_type type);
 
 /**
  * ib_bind_mw - Posts a work request to the send queue of the specified
@@ -2084,6 +2135,10 @@ struct ib_mw *ib_alloc_mw(struct ib_pd *pd);
  * @mw: The memory window to bind.
  * @mw_bind: Specifies information about the memory window, including
  *   its address range, remote access rights, and associated memory region.
+ *
+ * If there is no immediate error, the function will update the rkey member
+ * of the mw parameter to its new value. The bind operation can still fail
+ * asynchronously.
  */
 static inline int ib_bind_mw(struct ib_qp *qp,
 			     struct ib_mw *mw,
diff --git a/net/sunrpc/xprtrdma/verbs.c b/net/sunrpc/xprtrdma/verbs.c
index 745973b729af..93726560eaa8 100644
--- a/net/sunrpc/xprtrdma/verbs.c
+++ b/net/sunrpc/xprtrdma/verbs.c
@@ -1086,7 +1086,7 @@ rpcrdma_buffer_create(struct rpcrdma_buffer *buf, struct rpcrdma_ep *ep,
 	case RPCRDMA_MEMWINDOWS:
 		/* Allocate one extra request's worth, for full cycling */
 		for (i = (buf->rb_max_requests+1) * RPCRDMA_MAX_SEGS; i; i--) {
-			r->r.mw = ib_alloc_mw(ia->ri_pd);
+			r->r.mw = ib_alloc_mw(ia->ri_pd, IB_MW_TYPE_1);
 			if (IS_ERR(r->r.mw)) {
 				rc = PTR_ERR(r->r.mw);
 				dprintk("RPC:       %s: ib_alloc_mw"
@@ -1673,12 +1673,12 @@ rpcrdma_register_memwin_external(struct rpcrdma_mr_seg *seg,
 
 	*nsegs = 1;
 	rpcrdma_map_one(ia, seg, writing);
-	param.mr = ia->ri_bind_mem;
+	param.bind_info.mr = ia->ri_bind_mem;
 	param.wr_id = 0ULL;	/* no send cookie */
-	param.addr = seg->mr_dma;
-	param.length = seg->mr_len;
+	param.bind_info.addr = seg->mr_dma;
+	param.bind_info.length = seg->mr_len;
 	param.send_flags = 0;
-	param.mw_access_flags = mem_priv;
+	param.bind_info.mw_access_flags = mem_priv;
 
 	DECR_CQCOUNT(&r_xprt->rx_ep);
 	rc = ib_bind_mw(ia->ri_id->qp, seg->mr_chunk.rl_mw->r.mw, &param);
@@ -1690,7 +1690,7 @@ rpcrdma_register_memwin_external(struct rpcrdma_mr_seg *seg,
 		rpcrdma_unmap_one(ia, seg);
 	} else {
 		seg->mr_rkey = seg->mr_chunk.rl_mw->r.mw->rkey;
-		seg->mr_base = param.addr;
+		seg->mr_base = param.bind_info.addr;
 		seg->mr_nsegs = 1;
 	}
 	return rc;
@@ -1706,10 +1706,10 @@ rpcrdma_deregister_memwin_external(struct rpcrdma_mr_seg *seg,
 	int rc;
 
 	BUG_ON(seg->mr_nsegs != 1);
-	param.mr = ia->ri_bind_mem;
-	param.addr = 0ULL;	/* unbind */
-	param.length = 0;
-	param.mw_access_flags = 0;
+	param.bind_info.mr = ia->ri_bind_mem;
+	param.bind_info.addr = 0ULL;	/* unbind */
+	param.bind_info.length = 0;
+	param.bind_info.mw_access_flags = 0;
 	if (*r) {
 		param.wr_id = (u64) (unsigned long) *r;
 		param.send_flags = IB_SEND_SIGNALED;