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

* 'for-linus' of master.kernel.org:/pub/scm/linux/kernel/git/roland/infiniband:
  IB/iser: iSER Kconfig and Makefile
  IB/iser: iSER handling of memory for RDMA
  IB/iser: iSER RDMA CM (CMA) and IB verbs interaction
  IB/iser: iSER initiator iSCSI PDU and TX/RX
  IB/iser: iSCSI iSER transport provider high level code
  IB/iser: iSCSI iSER transport provider header file
  IB/uverbs: Remove unnecessary list_del()s
  IB/uverbs: Don't free wr list when it's known to be empty

drivers/infiniband/Kconfig

@@ -41,4 +41,6 @@ source "drivers/infiniband/ulp/ipoib/Kconfig"
 
 source "drivers/infiniband/ulp/srp/Kconfig"
 
+source "drivers/infiniband/ulp/iser/Kconfig"
+
 endmenu

drivers/infiniband/Makefile

@@ -3,3 +3,4 @@ obj-$(CONFIG_INFINIBAND_MTHCA)		+= hw/mthca/
 obj-$(CONFIG_IPATH_CORE)		+= hw/ipath/
 obj-$(CONFIG_INFINIBAND_IPOIB)		+= ulp/ipoib/
 obj-$(CONFIG_INFINIBAND_SRP)		+= ulp/srp/
+obj-$(CONFIG_INFINIBAND_ISER)		+= ulp/iser/

drivers/infiniband/core/uverbs_cmd.c

@@ -1530,7 +1530,6 @@ ssize_t ib_uverbs_post_send(struct ib_uverbs_file *file,
 out_put:
 	put_qp_read(qp);
 
-out:
 	while (wr) {
 		if (is_ud && wr->wr.ud.ah)
 			put_ah_read(wr->wr.ud.ah);
@@ -1539,6 +1538,7 @@ ssize_t ib_uverbs_post_send(struct ib_uverbs_file *file,
 		wr = next;
 	}
 
+out:
 	kfree(user_wr);
 
 	return ret ? ret : in_len;

drivers/infiniband/core/uverbs_main.c

@@ -188,7 +188,6 @@ static int ib_uverbs_cleanup_ucontext(struct ib_uverbs_file *file,
 
 		idr_remove_uobj(&ib_uverbs_ah_idr, uobj);
 		ib_destroy_ah(ah);
-		list_del(&uobj->list);
 		kfree(uobj);
 	}
 
@@ -200,7 +199,6 @@ static int ib_uverbs_cleanup_ucontext(struct ib_uverbs_file *file,
 		idr_remove_uobj(&ib_uverbs_qp_idr, uobj);
 		ib_uverbs_detach_umcast(qp, uqp);
 		ib_destroy_qp(qp);
-		list_del(&uobj->list);
 		ib_uverbs_release_uevent(file, &uqp->uevent);
 		kfree(uqp);
 	}
@@ -213,7 +211,6 @@ static int ib_uverbs_cleanup_ucontext(struct ib_uverbs_file *file,
 
 		idr_remove_uobj(&ib_uverbs_cq_idr, uobj);
 		ib_destroy_cq(cq);
-		list_del(&uobj->list);
 		ib_uverbs_release_ucq(file, ev_file, ucq);
 		kfree(ucq);
 	}
@@ -225,7 +222,6 @@ static int ib_uverbs_cleanup_ucontext(struct ib_uverbs_file *file,
 
 		idr_remove_uobj(&ib_uverbs_srq_idr, uobj);
 		ib_destroy_srq(srq);
-		list_del(&uobj->list);
 		ib_uverbs_release_uevent(file, uevent);
 		kfree(uevent);
 	}
@@ -243,7 +239,6 @@ static int ib_uverbs_cleanup_ucontext(struct ib_uverbs_file *file,
 		memobj = container_of(uobj, struct ib_umem_object, uobject);
 		ib_umem_release_on_close(mrdev, &memobj->umem);
 
-		list_del(&uobj->list);
 		kfree(memobj);
 	}
 
@@ -252,7 +247,6 @@ static int ib_uverbs_cleanup_ucontext(struct ib_uverbs_file *file,
 
 		idr_remove_uobj(&ib_uverbs_pd_idr, uobj);
 		ib_dealloc_pd(pd);
-		list_del(&uobj->list);
 		kfree(uobj);
 	}
 

drivers/infiniband/ulp/iser/Kconfig

+config INFINIBAND_ISER
+	tristate "ISCSI RDMA Protocol"
+	depends on INFINIBAND && SCSI
+	select SCSI_ISCSI_ATTRS
+	---help---
+	  Support for the ISCSI RDMA Protocol over InfiniBand.  This
+	  allows you to access storage devices that speak ISER/ISCSI
+	  over InfiniBand.
+
+	  The ISER protocol is defined by IETF.
+	  See <http://www.ietf.org/>.

drivers/infiniband/ulp/iser/Makefile

+obj-$(CONFIG_INFINIBAND_ISER)	+= ib_iser.o
+
+ib_iser-y			:= iser_verbs.o iser_initiator.o iser_memory.o \
+				   iscsi_iser.o

drivers/infiniband/ulp/iser/iscsi_iser.h

+/*
+ * iSER transport for the Open iSCSI Initiator & iSER transport internals
+ *
+ * Copyright (C) 2004 Dmitry Yusupov
+ * Copyright (C) 2004 Alex Aizman
+ * Copyright (C) 2005 Mike Christie
+ * based on code maintained by open-iscsi@googlegroups.com
+ *
+ * Copyright (c) 2004, 2005, 2006 Voltaire, Inc. All rights reserved.
+ * Copyright (c) 2005, 2006 Cisco Systems.  All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *	- Redistributions of source code must retain the above
+ *	  copyright notice, this list of conditions and the following
+ *	  disclaimer.
+ *
+ *	- Redistributions in binary form must reproduce the above
+ *	  copyright notice, this list of conditions and the following
+ *	  disclaimer in the documentation and/or other materials
+ *	  provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ *
+ * $Id: iscsi_iser.h 7051 2006-05-10 12:29:11Z ogerlitz $
+ */
+#ifndef __ISCSI_ISER_H__
+#define __ISCSI_ISER_H__
+
+#include <linux/types.h>
+#include <linux/net.h>
+#include <scsi/libiscsi.h>
+#include <scsi/scsi_transport_iscsi.h>
+
+#include <linux/wait.h>
+#include <linux/sched.h>
+#include <linux/list.h>
+#include <linux/slab.h>
+#include <linux/dma-mapping.h>
+#include <linux/mutex.h>
+#include <linux/mempool.h>
+#include <linux/uio.h>
+
+#include <linux/socket.h>
+#include <linux/in.h>
+#include <linux/in6.h>
+
+#include <rdma/ib_verbs.h>
+#include <rdma/ib_fmr_pool.h>
+#include <rdma/rdma_cm.h>
+
+#define DRV_NAME	"iser"
+#define PFX		DRV_NAME ": "
+#define DRV_VER		"0.1"
+#define DRV_DATE	"May 7th, 2006"
+
+#define iser_dbg(fmt, arg...)				\
+	do {						\
+		if (iser_debug_level > 0)		\
+			printk(KERN_DEBUG PFX "%s:" fmt,\
+				__func__ , ## arg);	\
+	} while (0)
+
+#define iser_err(fmt, arg...)				\
+	do {						\
+		printk(KERN_ERR PFX "%s:" fmt,          \
+		       __func__ , ## arg);		\
+	} while (0)
+
+					/* support upto 512KB in one RDMA */
+#define ISCSI_ISER_SG_TABLESIZE         (0x80000 >> PAGE_SHIFT)
+#define ISCSI_ISER_MAX_LUN		256
+#define ISCSI_ISER_MAX_CMD_LEN		16
+
+/* QP settings */
+/* Maximal bounds on received asynchronous PDUs */
+#define ISER_MAX_RX_MISC_PDUS		4 /* NOOP_IN(2) , ASYNC_EVENT(2)   */
+
+#define ISER_MAX_TX_MISC_PDUS		6 /* NOOP_OUT(2), TEXT(1),         *
+					   * SCSI_TMFUNC(2), LOGOUT(1) */
+
+#define ISER_QP_MAX_RECV_DTOS		(ISCSI_XMIT_CMDS_MAX + \
+					ISER_MAX_RX_MISC_PDUS    +  \
+					ISER_MAX_TX_MISC_PDUS)
+
+/* the max TX (send) WR supported by the iSER QP is defined by                 *
+ * max_send_wr = T * (1 + D) + C ; D is how many inflight dataouts we expect   *
+ * to have at max for SCSI command. The tx posting & completion handling code  *
+ * supports -EAGAIN scheme where tx is suspended till the QP has room for more *
+ * send WR. D=8 comes from 64K/8K                                              */
+
+#define ISER_INFLIGHT_DATAOUTS		8
+
+#define ISER_QP_MAX_REQ_DTOS		(ISCSI_XMIT_CMDS_MAX *    \
+					(1 + ISER_INFLIGHT_DATAOUTS) + \
+					ISER_MAX_TX_MISC_PDUS        + \
+					ISER_MAX_RX_MISC_PDUS)
+
+#define ISER_VER			0x10
+#define ISER_WSV			0x08
+#define ISER_RSV			0x04
+
+struct iser_hdr {
+	u8      flags;
+	u8      rsvd[3];
+	__be32  write_stag; /* write rkey */
+	__be64  write_va;
+	__be32  read_stag;  /* read rkey */
+	__be64  read_va;
+} __attribute__((packed));
+
+
+/* Length of an object name string */
+#define ISER_OBJECT_NAME_SIZE		    64
+
+enum iser_ib_conn_state {
+	ISER_CONN_INIT,		   /* descriptor allocd, no conn          */
+	ISER_CONN_PENDING,	   /* in the process of being established */
+	ISER_CONN_UP,		   /* up and running                      */
+	ISER_CONN_TERMINATING,	   /* in the process of being terminated  */
+	ISER_CONN_DOWN,		   /* shut down                           */
+	ISER_CONN_STATES_NUM
+};
+
+enum iser_task_status {
+	ISER_TASK_STATUS_INIT = 0,
+	ISER_TASK_STATUS_STARTED,
+	ISER_TASK_STATUS_COMPLETED
+};
+
+enum iser_data_dir {
+	ISER_DIR_IN = 0,	   /* to initiator */
+	ISER_DIR_OUT,		   /* from initiator */
+	ISER_DIRS_NUM
+};
+
+struct iser_data_buf {
+	void               *buf;      /* pointer to the sg list               */
+	unsigned int       size;      /* num entries of this sg               */
+	unsigned long      data_len;  /* total data len                       */
+	unsigned int       dma_nents; /* returned by dma_map_sg               */
+	char       	   *copy_buf; /* allocated copy buf for SGs unaligned *
+	                               * for rdma which are copied            */
+	struct scatterlist sg_single; /* SG-ified clone of a non SG SC or     *
+				       * unaligned SG                         */
+  };
+
+/* fwd declarations */
+struct iser_device;
+struct iscsi_iser_conn;
+struct iscsi_iser_cmd_task;
+
+struct iser_mem_reg {
+	u32  lkey;
+	u32  rkey;
+	u64  va;
+	u64  len;
+	void *mem_h;
+};
+
+struct iser_regd_buf {
+	struct iser_mem_reg     reg;        /* memory registration info        */
+	void                    *virt_addr;
+	struct iser_device      *device;    /* device->device for dma_unmap    */
+	dma_addr_t              dma_addr;   /* if non zero, addr for dma_unmap */
+	enum dma_data_direction direction;  /* direction for dma_unmap	       */
+	unsigned int            data_size;
+	atomic_t                ref_count;  /* refcount, freed when dec to 0   */
+};
+
+#define MAX_REGD_BUF_VECTOR_LEN	2
+
+struct iser_dto {
+	struct iscsi_iser_cmd_task *ctask;
+	struct iscsi_iser_conn     *conn;
+	int                        notify_enable;
+
+	/* vector of registered buffers */
+	unsigned int               regd_vector_len;
+	struct iser_regd_buf       *regd[MAX_REGD_BUF_VECTOR_LEN];
+
+	/* offset into the registered buffer may be specified */
+	unsigned int               offset[MAX_REGD_BUF_VECTOR_LEN];
+
+	/* a smaller size may be specified, if 0, then full size is used */
+	unsigned int               used_sz[MAX_REGD_BUF_VECTOR_LEN];
+};
+
+enum iser_desc_type {
+	ISCSI_RX,
+	ISCSI_TX_CONTROL ,
+	ISCSI_TX_SCSI_COMMAND,
+	ISCSI_TX_DATAOUT
+};
+
+struct iser_desc {
+	struct iser_hdr              iser_header;
+	struct iscsi_hdr             iscsi_header;
+	struct iser_regd_buf         hdr_regd_buf;
+	void                         *data;         /* used by RX & TX_CONTROL */
+	struct iser_regd_buf         data_regd_buf; /* used by RX & TX_CONTROL */
+	enum   iser_desc_type        type;
+	struct iser_dto              dto;
+};
+
+struct iser_device {
+	struct ib_device             *ib_device;
+	struct ib_pd	             *pd;
+	struct ib_cq	             *cq;
+	struct ib_mr	             *mr;
+	struct tasklet_struct	     cq_tasklet;
+	struct list_head             ig_list; /* entry in ig devices list */
+	int                          refcount;
+};
+
+struct iser_conn {
+	struct iscsi_iser_conn       *iser_conn; /* iser conn for upcalls  */
+	enum iser_ib_conn_state	     state;	    /* rdma connection state   */
+	spinlock_t		     lock;	    /* used for state changes  */
+	struct iser_device           *device;       /* device context          */
+	struct rdma_cm_id            *cma_id;       /* CMA ID		       */
+	struct ib_qp	             *qp;           /* QP 		       */
+	struct ib_fmr_pool           *fmr_pool;     /* pool of IB FMRs         */
+	int                          disc_evt_flag; /* disconn event delivered */
+	wait_queue_head_t	     wait;          /* waitq for conn/disconn  */
+	atomic_t                     post_recv_buf_count; /* posted rx count   */
+	atomic_t                     post_send_buf_count; /* posted tx count   */
+	struct work_struct           comperror_work; /* conn term sleepable ctx*/
+	char 			     name[ISER_OBJECT_NAME_SIZE];
+	struct iser_page_vec         *page_vec;     /* represents SG to fmr maps*
+						     * maps serialized as tx is*/
+	struct list_head	     conn_list;       /* entry in ig conn list */
+};
+
+struct iscsi_iser_conn {
+	struct iscsi_conn            *iscsi_conn;/* ptr to iscsi conn */
+	struct iser_conn             *ib_conn;   /* iSER IB conn      */
+
+	rwlock_t		     lock;
+};
+
+struct iscsi_iser_cmd_task {
+	struct iser_desc             desc;
+	struct iscsi_iser_conn	     *iser_conn;
+	int			     rdma_data_count;/* RDMA bytes           */
+	enum iser_task_status 	     status;
+	int                          command_sent;  /* set if command  sent  */
+	int                          dir[ISER_DIRS_NUM];      /* set if dir use*/
+	struct iser_regd_buf         rdma_regd[ISER_DIRS_NUM];/* regd rdma buf */
+	struct iser_data_buf         data[ISER_DIRS_NUM];     /* orig. data des*/
+	struct iser_data_buf         data_copy[ISER_DIRS_NUM];/* contig. copy  */
+};
+
+struct iser_page_vec {
+	u64 *pages;
+	int length;
+	int offset;
+	int data_size;
+};
+
+struct iser_global {
+	struct mutex      device_list_mutex;/*                   */
+	struct list_head  device_list;	     /* all iSER devices */
+	struct mutex      connlist_mutex;
+	struct list_head  connlist;		/* all iSER IB connections */
+
+	kmem_cache_t *desc_cache;
+};
+
+extern struct iser_global ig;
+extern int iser_debug_level;
+
+/* allocate connection resources needed for rdma functionality */
+int iser_conn_set_full_featured_mode(struct iscsi_conn *conn);
+
+int iser_send_control(struct iscsi_conn      *conn,
+		      struct iscsi_mgmt_task *mtask);
+
+int iser_send_command(struct iscsi_conn      *conn,
+		      struct iscsi_cmd_task  *ctask);
+
+int iser_send_data_out(struct iscsi_conn     *conn,
+		       struct iscsi_cmd_task *ctask,
+		       struct iscsi_data          *hdr);
+
+void iscsi_iser_recv(struct iscsi_conn *conn,
+		     struct iscsi_hdr       *hdr,
+		     char                   *rx_data,
+		     int                    rx_data_len);
+
+int  iser_conn_init(struct iser_conn **ib_conn);
+
+void iser_conn_terminate(struct iser_conn *ib_conn);
+
+void iser_conn_release(struct iser_conn *ib_conn);
+
+void iser_rcv_completion(struct iser_desc *desc,
+			 unsigned long    dto_xfer_len);
+
+void iser_snd_completion(struct iser_desc *desc);
+
+void iser_ctask_rdma_init(struct iscsi_iser_cmd_task     *ctask);
+
+void iser_ctask_rdma_finalize(struct iscsi_iser_cmd_task *ctask);
+
+void iser_dto_buffs_release(struct iser_dto *dto);
+
+int  iser_regd_buff_release(struct iser_regd_buf *regd_buf);
+
+void iser_reg_single(struct iser_device      *device,
+		     struct iser_regd_buf    *regd_buf,
+		     enum dma_data_direction direction);
+
+int  iser_start_rdma_unaligned_sg(struct iscsi_iser_cmd_task    *ctask,
+				  enum iser_data_dir            cmd_dir);
+
+void iser_finalize_rdma_unaligned_sg(struct iscsi_iser_cmd_task *ctask,
+				     enum iser_data_dir         cmd_dir);
+
+int  iser_reg_rdma_mem(struct iscsi_iser_cmd_task *ctask,
+		       enum   iser_data_dir        cmd_dir);
+
+int  iser_connect(struct iser_conn   *ib_conn,
+		  struct sockaddr_in *src_addr,
+		  struct sockaddr_in *dst_addr,
+		  int                non_blocking);
+
+int  iser_reg_page_vec(struct iser_conn     *ib_conn,
+		       struct iser_page_vec *page_vec,
+		       struct iser_mem_reg  *mem_reg);
+
+void iser_unreg_mem(struct iser_mem_reg *mem_reg);
+
+int  iser_post_recv(struct iser_desc *rx_desc);
+int  iser_post_send(struct iser_desc *tx_desc);
+
+int iser_conn_state_comp(struct iser_conn *ib_conn,
+			 enum iser_ib_conn_state comp);
+#endif

drivers/infiniband/ulp/iser/iser_memory.c

+/*
+ * Copyright (c) 2004, 2005, 2006 Voltaire, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *	- Redistributions of source code must retain the above
+ *	  copyright notice, this list of conditions and the following
+ *	  disclaimer.
+ *
+ *	- Redistributions in binary form must reproduce the above
+ *	  copyright notice, this list of conditions and the following
+ *	  disclaimer in the documentation and/or other materials
+ *	  provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ *
+ * $Id: iser_memory.c 6964 2006-05-07 11:11:43Z ogerlitz $
+ */
+#include <linux/module.h>
+#include <linux/kernel.h>
+#include <linux/slab.h>
+#include <linux/mm.h>
+#include <asm/io.h>
+#include <asm/scatterlist.h>
+#include <linux/scatterlist.h>
+
+#include "iscsi_iser.h"
+
+#define ISER_KMALLOC_THRESHOLD 0x20000 /* 128K - kmalloc limit */
+/**
+ * Decrements the reference count for the
+ * registered buffer & releases it
+ *
+ * returns 0 if released, 1 if deferred
+ */
+int iser_regd_buff_release(struct iser_regd_buf *regd_buf)
+{
+	struct device *dma_device;
+
+	if ((atomic_read(&regd_buf->ref_count) == 0) ||
+	    atomic_dec_and_test(&regd_buf->ref_count)) {
+		/* if we used the dma mr, unreg is just NOP */
+		if (regd_buf->reg.rkey != 0)
+			iser_unreg_mem(&regd_buf->reg);
+
+		if (regd_buf->dma_addr) {
+			dma_device = regd_buf->device->ib_device->dma_device;
+			dma_unmap_single(dma_device,
+					 regd_buf->dma_addr,
+					 regd_buf->data_size,
+					 regd_buf->direction);
+		}
+		/* else this regd buf is associated with task which we */
+		/* dma_unmap_single/sg later */
+		return 0;
+	} else {
+		iser_dbg("Release deferred, regd.buff: 0x%p\n", regd_buf);
+		return 1;
+	}
+}
+
+/**
+ * iser_reg_single - fills registered buffer descriptor with
+ *		     registration information
+ */
+void iser_reg_single(struct iser_device *device,
+		     struct iser_regd_buf *regd_buf,
+		     enum dma_data_direction direction)
+{
+	dma_addr_t dma_addr;
+
+	dma_addr  = dma_map_single(device->ib_device->dma_device,
+				   regd_buf->virt_addr,
+				   regd_buf->data_size, direction);
+	BUG_ON(dma_mapping_error(dma_addr));
+
+	regd_buf->reg.lkey = device->mr->lkey;
+	regd_buf->reg.rkey = 0; /* indicate there's no need to unreg */
+	regd_buf->reg.len  = regd_buf->data_size;
+	regd_buf->reg.va   = dma_addr;
+
+	regd_buf->dma_addr  = dma_addr;
+	regd_buf->direction = direction;
+}
+
+/**
+ * iser_start_rdma_unaligned_sg
+ */
+int iser_start_rdma_unaligned_sg(struct iscsi_iser_cmd_task  *iser_ctask,
+				 enum iser_data_dir cmd_dir)
+{
+	int dma_nents;
+	struct device *dma_device;
+	char *mem = NULL;
+	struct iser_data_buf *data = &iser_ctask->data[cmd_dir];
+	unsigned long  cmd_data_len = data->data_len;
+
+	if (cmd_data_len > ISER_KMALLOC_THRESHOLD)
+		mem = (void *)__get_free_pages(GFP_NOIO,
+		      long_log2(roundup_pow_of_two(cmd_data_len)) - PAGE_SHIFT);
+	else
+		mem = kmalloc(cmd_data_len, GFP_NOIO);
+
+	if (mem == NULL) {
+		iser_err("Failed to allocate mem size %d %d for copying sglist\n",
+			 data->size,(int)cmd_data_len);
+		return -ENOMEM;
+	}
+
+	if (cmd_dir == ISER_DIR_OUT) {
+		/* copy the unaligned sg the buffer which is used for RDMA */
+		struct scatterlist *sg = (struct scatterlist *)data->buf;
+		int i;
+		char *p, *from;
+
+		for (p = mem, i = 0; i < data->size; i++) {
+			from = kmap_atomic(sg[i].page, KM_USER0);
+			memcpy(p,
+			       from + sg[i].offset,
+			       sg[i].length);
+			kunmap_atomic(from, KM_USER0);
+			p += sg[i].length;
+		}
+	}
+
+	sg_init_one(&iser_ctask->data_copy[cmd_dir].sg_single, mem, cmd_data_len);
+	iser_ctask->data_copy[cmd_dir].buf  =
+		&iser_ctask->data_copy[cmd_dir].sg_single;
+	iser_ctask->data_copy[cmd_dir].size = 1;
+
+	iser_ctask->data_copy[cmd_dir].copy_buf  = mem;
+
+	dma_device = iser_ctask->iser_conn->ib_conn->device->ib_device->dma_device;
+
+	if (cmd_dir == ISER_DIR_OUT)
+		dma_nents = dma_map_sg(dma_device,
+				       &iser_ctask->data_copy[cmd_dir].sg_single,
+				       1, DMA_TO_DEVICE);
+	else
+		dma_nents = dma_map_sg(dma_device,
+				       &iser_ctask->data_copy[cmd_dir].sg_single,
+				       1, DMA_FROM_DEVICE);
+
+	BUG_ON(dma_nents == 0);
+
+	iser_ctask->data_copy[cmd_dir].dma_nents = dma_nents;
+	return 0;
+}
+
+/**
+ * iser_finalize_rdma_unaligned_sg
+ */
+void iser_finalize_rdma_unaligned_sg(struct iscsi_iser_cmd_task *iser_ctask,
+				     enum iser_data_dir         cmd_dir)
+{
+	struct device *dma_device;
+	struct iser_data_buf *mem_copy;
+	unsigned long  cmd_data_len;
+
+	dma_device = iser_ctask->iser_conn->ib_conn->device->ib_device->dma_device;
+	mem_copy   = &iser_ctask->data_copy[cmd_dir];
+
+	if (cmd_dir == ISER_DIR_OUT)
+		dma_unmap_sg(dma_device, &mem_copy->sg_single, 1,
+			     DMA_TO_DEVICE);
+	else
+		dma_unmap_sg(dma_device, &mem_copy->sg_single, 1,
+			     DMA_FROM_DEVICE);
+
+	if (cmd_dir == ISER_DIR_IN) {
+		char *mem;
+		struct scatterlist *sg;
+		unsigned char *p, *to;
+		unsigned int sg_size;
+		int i;
+
+		/* copy back read RDMA to unaligned sg */
+		mem	= mem_copy->copy_buf;
+
+		sg	= (struct scatterlist *)iser_ctask->data[ISER_DIR_IN].buf;
+		sg_size = iser_ctask->data[ISER_DIR_IN].size;
+
+		for (p = mem, i = 0; i < sg_size; i++){
+			to = kmap_atomic(sg[i].page, KM_SOFTIRQ0);
+			memcpy(to + sg[i].offset,
+			       p,
+			       sg[i].length);
+			kunmap_atomic(to, KM_SOFTIRQ0);
+			p += sg[i].length;
+		}
+	}
+
+	cmd_data_len = iser_ctask->data[cmd_dir].data_len;
+
+	if (cmd_data_len > ISER_KMALLOC_THRESHOLD)
+		free_pages((unsigned long)mem_copy->copy_buf,
+			   long_log2(roundup_pow_of_two(cmd_data_len)) - PAGE_SHIFT);
+	else
+		kfree(mem_copy->copy_buf);
+
+	mem_copy->copy_buf = NULL;
+}
+
+/**
+ * iser_sg_to_page_vec - Translates scatterlist entries to physical addresses
+ * and returns the length of resulting physical address array (may be less than
+ * the original due to possible compaction).
+ *
+ * we build a "page vec" under the assumption that the SG meets the RDMA
+ * alignment requirements. Other then the first and last SG elements, all
+ * the "internal" elements can be compacted into a list whose elements are
+ * dma addresses of physical pages. The code supports also the weird case
+ * where --few fragments of the same page-- are present in the SG as
+ * consecutive elements. Also, it handles one entry SG.
+ */
+static int iser_sg_to_page_vec(struct iser_data_buf *data,
+			       struct iser_page_vec *page_vec)
+{
+	struct scatterlist *sg = (struct scatterlist *)data->buf;
+	dma_addr_t first_addr, last_addr, page;
+	int start_aligned, end_aligned;
+	unsigned int cur_page = 0;
+	unsigned long total_sz = 0;
+	int i;
+
+	/* compute the offset of first element */
+	page_vec->offset = (u64) sg[0].offset;
+
+	for (i = 0; i < data->dma_nents; i++) {
+		total_sz += sg_dma_len(&sg[i]);
+
+		first_addr = sg_dma_address(&sg[i]);
+		last_addr  = first_addr + sg_dma_len(&sg[i]);
+
+		start_aligned = !(first_addr & ~PAGE_MASK);
+		end_aligned   = !(last_addr  & ~PAGE_MASK);
+
+		/* continue to collect page fragments till aligned or SG ends */
+		while (!end_aligned && (i + 1 < data->dma_nents)) {
+			i++;
+			total_sz += sg_dma_len(&sg[i]);
+			last_addr = sg_dma_address(&sg[i]) + sg_dma_len(&sg[i]);
+			end_aligned = !(last_addr  & ~PAGE_MASK);
+		}
+
+		first_addr = first_addr & PAGE_MASK;
+
+		for (page = first_addr; page < last_addr; page += PAGE_SIZE)
+			page_vec->pages[cur_page++] = page;
+
+	}
+	page_vec->data_size = total_sz;
+	iser_dbg("page_vec->data_size:%d cur_page %d\n", page_vec->data_size,cur_page);
+	return cur_page;
+}
+
+#define MASK_4K			((1UL << 12) - 1) /* 0xFFF */
+#define IS_4K_ALIGNED(addr)	((((unsigned long)addr) & MASK_4K) == 0)
+
+/**
+ * iser_data_buf_aligned_len - Tries to determine the maximal correctly aligned
+ * for RDMA sub-list of a scatter-gather list of memory buffers, and  returns
+ * the number of entries which are aligned correctly. Supports the case where
+ * consecutive SG elements are actually fragments of the same physcial page.
+ */
+static unsigned int iser_data_buf_aligned_len(struct iser_data_buf *data)
+{
+	struct scatterlist *sg;
+	dma_addr_t end_addr, next_addr;
+	int i, cnt;
+	unsigned int ret_len = 0;
+
+	sg = (struct scatterlist *)data->buf;
+
+	for (cnt = 0, i = 0; i < data->dma_nents; i++, cnt++) {
+		/* iser_dbg("Checking sg iobuf [%d]: phys=0x%08lX "
+		   "offset: %ld sz: %ld\n", i,
+		   (unsigned long)page_to_phys(sg[i].page),
+		   (unsigned long)sg[i].offset,
+		   (unsigned long)sg[i].length); */
+		end_addr = sg_dma_address(&sg[i]) +
+			   sg_dma_len(&sg[i]);
+		/* iser_dbg("Checking sg iobuf end address "
+		       "0x%08lX\n", end_addr); */
+		if (i + 1 < data->dma_nents) {
+			next_addr = sg_dma_address(&sg[i+1]);
+			/* are i, i+1 fragments of the same page? */
+			if (end_addr == next_addr)
+				continue;
+			else if (!IS_4K_ALIGNED(end_addr)) {
+				ret_len = cnt + 1;
+				break;
+			}
+		}
+	}
+	if (i == data->dma_nents)
+		ret_len = cnt;	/* loop ended */
+	iser_dbg("Found %d aligned entries out of %d in sg:0x%p\n",
+		 ret_len, data->dma_nents, data);
+	return ret_len;
+}
+
+static void iser_data_buf_dump(struct iser_data_buf *data)
+{
+	struct scatterlist *sg = (struct scatterlist *)data->buf;
+	int i;
+
+	for (i = 0; i < data->size; i++)
+		iser_err("sg[%d] dma_addr:0x%lX page:0x%p "
+			 "off:%d sz:%d dma_len:%d\n",
+			 i, (unsigned long)sg_dma_address(&sg[i]),
+			 sg[i].page, sg[i].offset,
+			 sg[i].length,sg_dma_len(&sg[i]));
+}
+
+static void iser_dump_page_vec(struct iser_page_vec *page_vec)
+{
+	int i;
+
+	iser_err("page vec length %d data size %d\n",
+		 page_vec->length, page_vec->data_size);
+	for (i = 0; i < page_vec->length; i++)
+		iser_err("%d %lx\n",i,(unsigned long)page_vec->pages[i]);
+}
+
+static void iser_page_vec_build(struct iser_data_buf *data,
+				struct iser_page_vec *page_vec)
+{
+	int page_vec_len = 0;
+
+	page_vec->length = 0;
+	page_vec->offset = 0;
+
+	iser_dbg("Translating sg sz: %d\n", data->dma_nents);
+	page_vec_len = iser_sg_to_page_vec(data,page_vec);
+	iser_dbg("sg len %d page_vec_len %d\n", data->dma_nents,page_vec_len);
+
+	page_vec->length = page_vec_len;
+
+	if (page_vec_len * PAGE_SIZE < page_vec->data_size) {
+		iser_err("page_vec too short to hold this SG\n");
+		iser_data_buf_dump(data);
+		iser_dump_page_vec(page_vec);
+		BUG();
+	}
+}
+
+/**
+ * iser_reg_rdma_mem - Registers memory intended for RDMA,
+ * obtaining rkey and va
+ *
+ * returns 0 on success, errno code on failure
+ */
+int iser_reg_rdma_mem(struct iscsi_iser_cmd_task *iser_ctask,
+		      enum   iser_data_dir        cmd_dir)
+{
+	struct iser_conn     *ib_conn = iser_ctask->iser_conn->ib_conn;
+	struct iser_data_buf *mem = &iser_ctask->data[cmd_dir];
+	struct iser_regd_buf *regd_buf;
+	int aligned_len;
+	int err;
+
+	regd_buf = &iser_ctask->rdma_regd[cmd_dir];
+
+	aligned_len = iser_data_buf_aligned_len(mem);
+	if (aligned_len != mem->size) {
+		iser_err("rdma alignment violation %d/%d aligned\n",
+			 aligned_len, mem->size);
+		iser_data_buf_dump(mem);
+		/* allocate copy buf, if we are writing, copy the */
+		/* unaligned scatterlist, dma map the copy        */
+		if (iser_start_rdma_unaligned_sg(iser_ctask, cmd_dir) != 0)
+				return -ENOMEM;
+		mem = &iser_ctask->data_copy[cmd_dir];
+	}
+
+	iser_page_vec_build(mem, ib_conn->page_vec);
+	err = iser_reg_page_vec(ib_conn, ib_conn->page_vec, &regd_buf->reg);
+	if (err)
+		return err;
+
+	/* take a reference on this regd buf such that it will not be released *
+	 * (eg in send dto completion) before we get the scsi response         */
+	atomic_inc(&regd_buf->ref_count);
+	return 0;
+}

drivers/scsi/Makefile

@@ -34,6 +34,7 @@ obj-$(CONFIG_SCSI_ISCSI_ATTRS)	+= scsi_transport_iscsi.o
 obj-$(CONFIG_SCSI_SAS_ATTRS)	+= scsi_transport_sas.o
 
 obj-$(CONFIG_ISCSI_TCP) 	+= libiscsi.o	iscsi_tcp.o
+obj-$(CONFIG_INFINIBAND_ISER) 	+= libiscsi.o
 obj-$(CONFIG_SCSI_AMIGA7XX)	+= amiga7xx.o	53c7xx.o
 obj-$(CONFIG_A3000_SCSI)	+= a3000.o	wd33c93.o
 obj-$(CONFIG_A2091_SCSI)	+= a2091.o	wd33c93.o