[SCSI] scsi tgt: SCSI RDMA Protocol library functions

libsrp provides helper functions for SRP target drivers.

Some SRP target drivers would be out of drivers/scsi/ so we added an
entry for libsrp in drivers/scsi/Kconfig.

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: Santiago Leon <santil@us.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

drivers/scsi/Kconfig

@@ -1760,6 +1760,16 @@ config ZFCP
           called zfcp. If you want to compile it as a module, say M here
           and read <file:Documentation/modules.txt>.
 
+config SCSI_SRP
+	tristate "SCSI RDMA Protocol helper library"
+	depends on SCSI && PCI
+	select SCSI_TGT
+	help
+	  If you wish to use SRP target drivers, say Y.
+
+	  To compile this driver as a module, choose M here: the
+	  module will be called libsrp.
+
 endmenu
 
 source "drivers/scsi/pcmcia/Kconfig"

drivers/scsi/Makefile

@@ -126,6 +126,7 @@ obj-$(CONFIG_SCSI_FCAL)		+= fcal.o
 obj-$(CONFIG_SCSI_LASI700)	+= 53c700.o lasi700.o
 obj-$(CONFIG_SCSI_NSP32)	+= nsp32.o
 obj-$(CONFIG_SCSI_IPR)		+= ipr.o
+obj-$(CONFIG_SCSI_SRP)		+= libsrp.o
 obj-$(CONFIG_SCSI_IBMVSCSI)	+= ibmvscsi/
 obj-$(CONFIG_SCSI_HPTIOP)	+= hptiop.o
 obj-$(CONFIG_SCSI_STEX)		+= stex.o

drivers/scsi/libsrp.c

@@ -0,0 +1,441 @@
+/*
+ * SCSI RDAM Protocol lib functions
+ *
+ * Copyright (C) 2006 FUJITA Tomonori <tomof@acm.org>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation; either version 2 of the
+ * License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA
+ * 02110-1301 USA
+ */
+#include <linux/err.h>
+#include <linux/kfifo.h>
+#include <linux/scatterlist.h>
+#include <linux/dma-mapping.h>
+#include <linux/pci.h>
+#include <scsi/scsi.h>
+#include <scsi/scsi_cmnd.h>
+#include <scsi/scsi_tcq.h>
+#include <scsi/scsi_tgt.h>
+#include <scsi/srp.h>
+#include <scsi/libsrp.h>
+
+enum srp_task_attributes {
+	SRP_SIMPLE_TASK = 0,
+	SRP_HEAD_TASK = 1,
+	SRP_ORDERED_TASK = 2,
+	SRP_ACA_TASK = 4
+};
+
+/* tmp - will replace with SCSI logging stuff */
+#define eprintk(fmt, args...)					\
+do {								\
+	printk("%s(%d) " fmt, __FUNCTION__, __LINE__, ##args);	\
+} while (0)
+/* #define dprintk eprintk */
+#define dprintk(fmt, args...)
+
+static int srp_iu_pool_alloc(struct srp_queue *q, size_t max,
+			     struct srp_buf **ring)
+{
+	int i;
+	struct iu_entry *iue;
+
+	q->pool = kcalloc(max, sizeof(struct iu_entry *), GFP_KERNEL);
+	if (!q->pool)
+		return -ENOMEM;
+	q->items = kcalloc(max, sizeof(struct iu_entry), GFP_KERNEL);
+	if (!q->items)
+		goto free_pool;
+
+	spin_lock_init(&q->lock);
+	q->queue = kfifo_init((void *) q->pool, max * sizeof(void *),
+			      GFP_KERNEL, &q->lock);
+	if (IS_ERR(q->queue))
+		goto free_item;
+
+	for (i = 0, iue = q->items; i < max; i++) {
+		__kfifo_put(q->queue, (void *) &iue, sizeof(void *));
+		iue->sbuf = ring[i];
+		iue++;
+	}
+	return 0;
+
+free_item:
+	kfree(q->items);
+free_pool:
+	kfree(q->pool);
+	return -ENOMEM;
+}
+
+static void srp_iu_pool_free(struct srp_queue *q)
+{
+	kfree(q->items);
+	kfree(q->pool);
+}
+
+static struct srp_buf **srp_ring_alloc(struct device *dev,
+				       size_t max, size_t size)
+{
+	int i;
+	struct srp_buf **ring;
+
+	ring = kcalloc(max, sizeof(struct srp_buf *), GFP_KERNEL);
+	if (!ring)
+		return NULL;
+
+	for (i = 0; i < max; i++) {
+		ring[i] = kzalloc(sizeof(struct srp_buf), GFP_KERNEL);
+		if (!ring[i])
+			goto out;
+		ring[i]->buf = dma_alloc_coherent(dev, size, &ring[i]->dma,
+						  GFP_KERNEL);
+		if (!ring[i]->buf)
+			goto out;
+	}
+	return ring;
+
+out:
+	for (i = 0; i < max && ring[i]; i++) {
+		if (ring[i]->buf)
+			dma_free_coherent(dev, size, ring[i]->buf, ring[i]->dma);
+		kfree(ring[i]);
+	}
+	kfree(ring);
+
+	return NULL;
+}
+
+static void srp_ring_free(struct device *dev, struct srp_buf **ring, size_t max,
+			  size_t size)
+{
+	int i;
+
+	for (i = 0; i < max; i++) {
+		dma_free_coherent(dev, size, ring[i]->buf, ring[i]->dma);
+		kfree(ring[i]);
+	}
+}
+
+int srp_target_alloc(struct srp_target *target, struct device *dev,
+		     size_t nr, size_t iu_size)
+{
+	int err;
+
+	spin_lock_init(&target->lock);
+	INIT_LIST_HEAD(&target->cmd_queue);
+
+	target->dev = dev;
+	target->dev->driver_data = target;
+
+	target->srp_iu_size = iu_size;
+	target->rx_ring_size = nr;
+	target->rx_ring = srp_ring_alloc(target->dev, nr, iu_size);
+	if (!target->rx_ring)
+		return -ENOMEM;
+	err = srp_iu_pool_alloc(&target->iu_queue, nr, target->rx_ring);
+	if (err)
+		goto free_ring;
+
+	return 0;
+
+free_ring:
+	srp_ring_free(target->dev, target->rx_ring, nr, iu_size);
+	return -ENOMEM;
+}
+EXPORT_SYMBOL_GPL(srp_target_alloc);
+
+void srp_target_free(struct srp_target *target)
+{
+	srp_ring_free(target->dev, target->rx_ring, target->rx_ring_size,
+		      target->srp_iu_size);
+	srp_iu_pool_free(&target->iu_queue);
+}
+EXPORT_SYMBOL_GPL(srp_target_free);
+
+struct iu_entry *srp_iu_get(struct srp_target *target)
+{
+	struct iu_entry *iue = NULL;
+
+	kfifo_get(target->iu_queue.queue, (void *) &iue, sizeof(void *));
+	if (!iue)
+		return iue;
+	iue->target = target;
+	INIT_LIST_HEAD(&iue->ilist);
+	iue->flags = 0;
+	return iue;
+}
+EXPORT_SYMBOL_GPL(srp_iu_get);
+
+void srp_iu_put(struct iu_entry *iue)
+{
+	kfifo_put(iue->target->iu_queue.queue, (void *) &iue, sizeof(void *));
+}
+EXPORT_SYMBOL_GPL(srp_iu_put);
+
+static int srp_direct_data(struct scsi_cmnd *sc, struct srp_direct_buf *md,
+			   enum dma_data_direction dir, srp_rdma_t rdma_io,
+			   int dma_map, int ext_desc)
+{
+	struct iu_entry *iue = NULL;
+	struct scatterlist *sg = NULL;
+	int err, nsg = 0, len;
+
+	if (dma_map) {
+		iue = (struct iu_entry *) sc->SCp.ptr;
+		sg = sc->request_buffer;
+
+		dprintk("%p %u %u %d\n", iue, sc->request_bufflen,
+			md->len, sc->use_sg);
+
+		nsg = dma_map_sg(iue->target->dev, sg, sc->use_sg,
+				 DMA_BIDIRECTIONAL);
+		if (!nsg) {
+			printk("fail to map %p %d\n", iue, sc->use_sg);
+			return 0;
+		}
+		len = min(sc->request_bufflen, md->len);
+	} else
+		len = md->len;
+
+	err = rdma_io(sc, sg, nsg, md, 1, dir, len);
+
+	if (dma_map)
+		dma_unmap_sg(iue->target->dev, sg, nsg, DMA_BIDIRECTIONAL);
+
+	return err;
+}
+
+static int srp_indirect_data(struct scsi_cmnd *sc, struct srp_cmd *cmd,
+			     struct srp_indirect_buf *id,
+			     enum dma_data_direction dir, srp_rdma_t rdma_io,
+			     int dma_map, int ext_desc)
+{
+	struct iu_entry *iue = NULL;
+	struct srp_direct_buf *md = NULL;
+	struct scatterlist dummy, *sg = NULL;
+	dma_addr_t token = 0;
+	long err;
+	unsigned int done = 0;
+	int nmd, nsg = 0, len;
+
+	if (dma_map || ext_desc) {
+		iue = (struct iu_entry *) sc->SCp.ptr;
+		sg = sc->request_buffer;
+
+		dprintk("%p %u %u %d %d\n",
+			iue, sc->request_bufflen, id->len,
+			cmd->data_in_desc_cnt, cmd->data_out_desc_cnt);
+	}
+
+	nmd = id->table_desc.len / sizeof(struct srp_direct_buf);
+
+	if ((dir == DMA_FROM_DEVICE && nmd == cmd->data_in_desc_cnt) ||
+	    (dir == DMA_TO_DEVICE && nmd == cmd->data_out_desc_cnt)) {
+		md = &id->desc_list[0];
+		goto rdma;
+	}
+
+	if (ext_desc && dma_map) {
+		md = dma_alloc_coherent(iue->target->dev, id->table_desc.len,
+				&token, GFP_KERNEL);
+		if (!md) {
+			eprintk("Can't get dma memory %u\n", id->table_desc.len);
+			return -ENOMEM;
+		}
+
+		sg_init_one(&dummy, md, id->table_desc.len);
+		sg_dma_address(&dummy) = token;
+		err = rdma_io(sc, &dummy, 1, &id->table_desc, 1, DMA_TO_DEVICE,
+			      id->table_desc.len);
+		if (err < 0) {
+			eprintk("Error copying indirect table %ld\n", err);
+			goto free_mem;
+		}
+	} else {
+		eprintk("This command uses external indirect buffer\n");
+		return -EINVAL;
+	}
+
+rdma:
+	if (dma_map) {
+		nsg = dma_map_sg(iue->target->dev, sg, sc->use_sg, DMA_BIDIRECTIONAL);
+		if (!nsg) {
+			eprintk("fail to map %p %d\n", iue, sc->use_sg);
+			goto free_mem;
+		}
+		len = min(sc->request_bufflen, id->len);
+	} else
+		len = id->len;
+
+	err = rdma_io(sc, sg, nsg, md, nmd, dir, len);
+
+	if (dma_map)
+		dma_unmap_sg(iue->target->dev, sg, nsg, DMA_BIDIRECTIONAL);
+
+free_mem:
+	if (token && dma_map)
+		dma_free_coherent(iue->target->dev, id->table_desc.len, md, token);
+
+	return done;
+}
+
+static int data_out_desc_size(struct srp_cmd *cmd)
+{
+	int size = 0;
+	u8 fmt = cmd->buf_fmt >> 4;
+
+	switch (fmt) {
+	case SRP_NO_DATA_DESC:
+		break;
+	case SRP_DATA_DESC_DIRECT:
+		size = sizeof(struct srp_direct_buf);
+		break;
+	case SRP_DATA_DESC_INDIRECT:
+		size = sizeof(struct srp_indirect_buf) +
+			sizeof(struct srp_direct_buf) * cmd->data_out_desc_cnt;
+		break;
+	default:
+		eprintk("client error. Invalid data_out_format %x\n", fmt);
+		break;
+	}
+	return size;
+}
+
+/*
+ * TODO: this can be called multiple times for a single command if it
+ * has very long data.
+ */
+int srp_transfer_data(struct scsi_cmnd *sc, struct srp_cmd *cmd,
+		      srp_rdma_t rdma_io, int dma_map, int ext_desc)
+{
+	struct srp_direct_buf *md;
+	struct srp_indirect_buf *id;
+	enum dma_data_direction dir;
+	int offset, err = 0;
+	u8 format;
+
+	offset = cmd->add_cdb_len * 4;
+
+	dir = srp_cmd_direction(cmd);
+	if (dir == DMA_FROM_DEVICE)
+		offset += data_out_desc_size(cmd);
+
+	if (dir == DMA_TO_DEVICE)
+		format = cmd->buf_fmt >> 4;
+	else
+		format = cmd->buf_fmt & ((1U << 4) - 1);
+
+	switch (format) {
+	case SRP_NO_DATA_DESC:
+		break;
+	case SRP_DATA_DESC_DIRECT:
+		md = (struct srp_direct_buf *)
+			(cmd->add_data + offset);
+		err = srp_direct_data(sc, md, dir, rdma_io, dma_map, ext_desc);
+		break;
+	case SRP_DATA_DESC_INDIRECT:
+		id = (struct srp_indirect_buf *)
+			(cmd->add_data + offset);
+		err = srp_indirect_data(sc, cmd, id, dir, rdma_io, dma_map,
+					ext_desc);
+		break;
+	default:
+		eprintk("Unknown format %d %x\n", dir, format);
+		break;
+	}
+
+	return err;
+}
+EXPORT_SYMBOL_GPL(srp_transfer_data);
+
+static int vscsis_data_length(struct srp_cmd *cmd, enum dma_data_direction dir)
+{
+	struct srp_direct_buf *md;
+	struct srp_indirect_buf *id;
+	int len = 0, offset = cmd->add_cdb_len * 4;
+	u8 fmt;
+
+	if (dir == DMA_TO_DEVICE)
+		fmt = cmd->buf_fmt >> 4;
+	else {
+		fmt = cmd->buf_fmt & ((1U << 4) - 1);
+		offset += data_out_desc_size(cmd);
+	}
+
+	switch (fmt) {
+	case SRP_NO_DATA_DESC:
+		break;
+	case SRP_DATA_DESC_DIRECT:
+		md = (struct srp_direct_buf *) (cmd->add_data + offset);
+		len = md->len;
+		break;
+	case SRP_DATA_DESC_INDIRECT:
+		id = (struct srp_indirect_buf *) (cmd->add_data + offset);
+		len = id->len;
+		break;
+	default:
+		eprintk("invalid data format %x\n", fmt);
+		break;
+	}
+	return len;
+}
+
+int srp_cmd_queue(struct Scsi_Host *shost, struct srp_cmd *cmd, void *info,
+		  u64 addr)
+{
+	enum dma_data_direction dir;
+	struct scsi_cmnd *sc;
+	int tag, len, err;
+
+	switch (cmd->task_attr) {
+	case SRP_SIMPLE_TASK:
+		tag = MSG_SIMPLE_TAG;
+		break;
+	case SRP_ORDERED_TASK:
+		tag = MSG_ORDERED_TAG;
+		break;
+	case SRP_HEAD_TASK:
+		tag = MSG_HEAD_TAG;
+		break;
+	default:
+		eprintk("Task attribute %d not supported\n", cmd->task_attr);
+		tag = MSG_ORDERED_TAG;
+	}
+
+	dir = srp_cmd_direction(cmd);
+	len = vscsis_data_length(cmd, dir);
+
+	dprintk("%p %x %lx %d %d %d %llx\n", info, cmd->cdb[0],
+		cmd->lun, dir, len, tag, (unsigned long long) cmd->tag);
+
+	sc = scsi_host_get_command(shost, dir, GFP_KERNEL);
+	if (!sc)
+		return -ENOMEM;
+
+	sc->SCp.ptr = info;
+	memcpy(sc->cmnd, cmd->cdb, MAX_COMMAND_SIZE);
+	sc->request_bufflen = len;
+	sc->request_buffer = (void *) (unsigned long) addr;
+	sc->tag = tag;
+	err = scsi_tgt_queue_command(sc, (struct scsi_lun *) &cmd->lun, cmd->tag);
+	if (err)
+		scsi_host_put_command(shost, sc);
+
+	return err;
+}
+EXPORT_SYMBOL_GPL(srp_cmd_queue);
+
+MODULE_DESCRIPTION("SCSI RDAM Protocol lib functions");
+MODULE_AUTHOR("FUJITA Tomonori");
+MODULE_LICENSE("GPL");

include/scsi/libsrp.h

@@ -0,0 +1,77 @@
+#ifndef __LIBSRP_H__
+#define __LIBSRP_H__
+
+#include <linux/list.h>
+#include <scsi/scsi_cmnd.h>
+#include <scsi/scsi_host.h>
+#include <scsi/srp.h>
+
+enum iue_flags {
+	V_DIOVER,
+	V_WRITE,
+	V_LINKED,
+	V_FLYING,
+};
+
+struct srp_buf {
+	dma_addr_t dma;
+	void *buf;
+};
+
+struct srp_queue {
+	void *pool;
+	void *items;
+	struct kfifo *queue;
+	spinlock_t lock;
+};
+
+struct srp_target {
+	struct Scsi_Host *shost;
+	struct device *dev;
+
+	spinlock_t lock;
+	struct list_head cmd_queue;
+
+	size_t srp_iu_size;
+	struct srp_queue iu_queue;
+	size_t rx_ring_size;
+	struct srp_buf **rx_ring;
+
+	void *ldata;
+};
+
+struct iu_entry {
+	struct srp_target *target;
+
+	struct list_head ilist;
+	dma_addr_t remote_token;
+	unsigned long flags;
+
+	struct srp_buf *sbuf;
+};
+
+typedef int (srp_rdma_t)(struct scsi_cmnd *, struct scatterlist *, int,
+			 struct srp_direct_buf *, int,
+			 enum dma_data_direction, unsigned int);
+extern int srp_target_alloc(struct srp_target *, struct device *, size_t, size_t);
+extern void srp_target_free(struct srp_target *);
+
+extern struct iu_entry *srp_iu_get(struct srp_target *);
+extern void srp_iu_put(struct iu_entry *);
+
+extern int srp_cmd_queue(struct Scsi_Host *, struct srp_cmd *, void *, u64);
+extern int srp_transfer_data(struct scsi_cmnd *, struct srp_cmd *,
+			     srp_rdma_t, int, int);
+
+
+static inline struct srp_target *host_to_srp_target(struct Scsi_Host *host)
+{
+	return (struct srp_target *) host->hostdata;
+}
+
+static inline int srp_cmd_direction(struct srp_cmd *cmd)
+{
+	return (cmd->buf_fmt >> 4) ? DMA_TO_DEVICE : DMA_FROM_DEVICE;
+}
+
+#endif