blob: 06f578cde75b002bbc935e45516d3bf1266c2ca3 [file] [log] [blame]
Or Gerlitz49cd5382006-05-11 10:00:21 +03001/*
2 * iSER transport for the Open iSCSI Initiator & iSER transport internals
3 *
4 * Copyright (C) 2004 Dmitry Yusupov
5 * Copyright (C) 2004 Alex Aizman
6 * Copyright (C) 2005 Mike Christie
7 * based on code maintained by open-iscsi@googlegroups.com
8 *
9 * Copyright (c) 2004, 2005, 2006 Voltaire, Inc. All rights reserved.
10 * Copyright (c) 2005, 2006 Cisco Systems. All rights reserved.
11 *
12 * This software is available to you under a choice of one of two
13 * licenses. You may choose to be licensed under the terms of the GNU
14 * General Public License (GPL) Version 2, available from the file
15 * COPYING in the main directory of this source tree, or the
16 * OpenIB.org BSD license below:
17 *
18 * Redistribution and use in source and binary forms, with or
19 * without modification, are permitted provided that the following
20 * conditions are met:
21 *
22 * - Redistributions of source code must retain the above
23 * copyright notice, this list of conditions and the following
24 * disclaimer.
25 *
26 * - Redistributions in binary form must reproduce the above
27 * copyright notice, this list of conditions and the following
28 * disclaimer in the documentation and/or other materials
29 * provided with the distribution.
30 *
31 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
32 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
33 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
34 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
35 * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
36 * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
37 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
38 * SOFTWARE.
Or Gerlitz49cd5382006-05-11 10:00:21 +030039 */
40#ifndef __ISCSI_ISER_H__
41#define __ISCSI_ISER_H__
42
43#include <linux/types.h>
44#include <linux/net.h>
Roi Dayan4f363882013-05-01 13:25:25 +000045#include <linux/printk.h>
Or Gerlitz49cd5382006-05-11 10:00:21 +030046#include <scsi/libiscsi.h>
47#include <scsi/scsi_transport_iscsi.h>
48
Alexey Dobriyana6b7a402011-06-06 10:43:46 +000049#include <linux/interrupt.h>
Or Gerlitz49cd5382006-05-11 10:00:21 +030050#include <linux/wait.h>
51#include <linux/sched.h>
52#include <linux/list.h>
53#include <linux/slab.h>
54#include <linux/dma-mapping.h>
55#include <linux/mutex.h>
56#include <linux/mempool.h>
57#include <linux/uio.h>
58
59#include <linux/socket.h>
60#include <linux/in.h>
61#include <linux/in6.h>
62
63#include <rdma/ib_verbs.h>
64#include <rdma/ib_fmr_pool.h>
65#include <rdma/rdma_cm.h>
66
67#define DRV_NAME "iser"
68#define PFX DRV_NAME ": "
Roi Dayanc1d786e2013-05-01 13:25:24 +000069#define DRV_VER "1.1"
Or Gerlitz49cd5382006-05-11 10:00:21 +030070
71#define iser_dbg(fmt, arg...) \
72 do { \
Roi Dayan4f363882013-05-01 13:25:25 +000073 if (iser_debug_level > 2) \
Eli Dorfman6f735e32008-04-29 13:46:52 -070074 printk(KERN_DEBUG PFX "%s:" fmt,\
75 __func__ , ## arg); \
76 } while (0)
77
78#define iser_warn(fmt, arg...) \
79 do { \
Roi Dayan4f363882013-05-01 13:25:25 +000080 if (iser_debug_level > 1) \
81 pr_warn(PFX "%s:" fmt, \
82 __func__ , ## arg); \
83 } while (0)
84
85#define iser_info(fmt, arg...) \
86 do { \
Or Gerlitz49cd5382006-05-11 10:00:21 +030087 if (iser_debug_level > 0) \
Roi Dayan4f363882013-05-01 13:25:25 +000088 pr_info(PFX "%s:" fmt, \
Or Gerlitz49cd5382006-05-11 10:00:21 +030089 __func__ , ## arg); \
90 } while (0)
91
92#define iser_err(fmt, arg...) \
93 do { \
94 printk(KERN_ERR PFX "%s:" fmt, \
95 __func__ , ## arg); \
96 } while (0)
97
Erez Zilber8dfa0872006-09-11 12:22:30 +030098#define SHIFT_4K 12
Or Gerlitz0ace64b2011-08-01 21:12:09 +000099#define SIZE_4K (1ULL << SHIFT_4K)
Erez Zilber8dfa0872006-09-11 12:22:30 +0300100#define MASK_4K (~(SIZE_4K-1))
101
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300102 /* support up to 512KB in one RDMA */
Erez Zilber8dfa0872006-09-11 12:22:30 +0300103#define ISCSI_ISER_SG_TABLESIZE (0x80000 >> SHIFT_4K)
Or Gerlitzb96e4ab2013-02-21 14:50:08 +0000104#define ISER_DEF_CMD_PER_LUN ISCSI_DEF_XMIT_CMDS_MAX
Or Gerlitz49cd5382006-05-11 10:00:21 +0300105
106/* QP settings */
107/* Maximal bounds on received asynchronous PDUs */
108#define ISER_MAX_RX_MISC_PDUS 4 /* NOOP_IN(2) , ASYNC_EVENT(2) */
109
110#define ISER_MAX_TX_MISC_PDUS 6 /* NOOP_OUT(2), TEXT(1), *
111 * SCSI_TMFUNC(2), LOGOUT(1) */
112
Or Gerlitzbcc60c32010-02-08 13:17:42 +0000113#define ISER_QP_MAX_RECV_DTOS (ISCSI_DEF_XMIT_CMDS_MAX)
114
115#define ISER_MIN_POSTED_RX (ISCSI_DEF_XMIT_CMDS_MAX >> 2)
Or Gerlitz49cd5382006-05-11 10:00:21 +0300116
117/* the max TX (send) WR supported by the iSER QP is defined by *
118 * max_send_wr = T * (1 + D) + C ; D is how many inflight dataouts we expect *
119 * to have at max for SCSI command. The tx posting & completion handling code *
120 * supports -EAGAIN scheme where tx is suspended till the QP has room for more *
121 * send WR. D=8 comes from 64K/8K */
122
123#define ISER_INFLIGHT_DATAOUTS 8
124
Mike Christie15482712007-05-30 12:57:19 -0500125#define ISER_QP_MAX_REQ_DTOS (ISCSI_DEF_XMIT_CMDS_MAX * \
Or Gerlitz49cd5382006-05-11 10:00:21 +0300126 (1 + ISER_INFLIGHT_DATAOUTS) + \
127 ISER_MAX_TX_MISC_PDUS + \
128 ISER_MAX_RX_MISC_PDUS)
129
130#define ISER_VER 0x10
131#define ISER_WSV 0x08
132#define ISER_RSV 0x04
133
134struct iser_hdr {
135 u8 flags;
136 u8 rsvd[3];
137 __be32 write_stag; /* write rkey */
138 __be64 write_va;
139 __be32 read_stag; /* read rkey */
140 __be64 read_va;
141} __attribute__((packed));
142
Or Gerlitz8d8399d2013-05-01 13:25:27 +0000143
144#define ISER_ZBVA_NOT_SUPPORTED 0x80
145#define ISER_SEND_W_INV_NOT_SUPPORTED 0x40
146
147struct iser_cm_hdr {
148 u8 flags;
149 u8 rsvd[3];
150} __packed;
151
Or Gerlitzbcc60c32010-02-08 13:17:42 +0000152/* Constant PDU lengths calculations */
153#define ISER_HEADERS_LEN (sizeof(struct iser_hdr) + sizeof(struct iscsi_hdr))
154
155#define ISER_RECV_DATA_SEG_LEN 128
156#define ISER_RX_PAYLOAD_SIZE (ISER_HEADERS_LEN + ISER_RECV_DATA_SEG_LEN)
157#define ISER_RX_LOGIN_SIZE (ISER_HEADERS_LEN + ISCSI_DEF_MAX_RECV_SEG_LEN)
Or Gerlitz49cd5382006-05-11 10:00:21 +0300158
159/* Length of an object name string */
160#define ISER_OBJECT_NAME_SIZE 64
161
162enum iser_ib_conn_state {
163 ISER_CONN_INIT, /* descriptor allocd, no conn */
164 ISER_CONN_PENDING, /* in the process of being established */
165 ISER_CONN_UP, /* up and running */
166 ISER_CONN_TERMINATING, /* in the process of being terminated */
167 ISER_CONN_DOWN, /* shut down */
168 ISER_CONN_STATES_NUM
169};
170
171enum iser_task_status {
172 ISER_TASK_STATUS_INIT = 0,
173 ISER_TASK_STATUS_STARTED,
174 ISER_TASK_STATUS_COMPLETED
175};
176
177enum iser_data_dir {
178 ISER_DIR_IN = 0, /* to initiator */
179 ISER_DIR_OUT, /* from initiator */
180 ISER_DIRS_NUM
181};
182
183struct iser_data_buf {
184 void *buf; /* pointer to the sg list */
185 unsigned int size; /* num entries of this sg */
186 unsigned long data_len; /* total data len */
187 unsigned int dma_nents; /* returned by dma_map_sg */
188 char *copy_buf; /* allocated copy buf for SGs unaligned *
189 * for rdma which are copied */
190 struct scatterlist sg_single; /* SG-ified clone of a non SG SC or *
191 * unaligned SG */
192 };
193
194/* fwd declarations */
195struct iser_device;
Alex Tabachnik5a33a662012-09-23 15:17:44 +0000196struct iser_cq_desc;
Or Gerlitz49cd5382006-05-11 10:00:21 +0300197struct iscsi_iser_conn;
Mike Christie2261ec32008-05-21 15:54:11 -0500198struct iscsi_iser_task;
Mike Christie412eeaf2008-05-21 15:54:14 -0500199struct iscsi_endpoint;
Or Gerlitz49cd5382006-05-11 10:00:21 +0300200
201struct iser_mem_reg {
202 u32 lkey;
203 u32 rkey;
204 u64 va;
205 u64 len;
206 void *mem_h;
Erez Zilberd8111022006-09-11 12:26:33 +0300207 int is_fmr;
Or Gerlitz49cd5382006-05-11 10:00:21 +0300208};
209
210struct iser_regd_buf {
211 struct iser_mem_reg reg; /* memory registration info */
212 void *virt_addr;
213 struct iser_device *device; /* device->device for dma_unmap */
Or Gerlitz49cd5382006-05-11 10:00:21 +0300214 enum dma_data_direction direction; /* direction for dma_unmap */
215 unsigned int data_size;
Or Gerlitz49cd5382006-05-11 10:00:21 +0300216};
217
218enum iser_desc_type {
Or Gerlitz49cd5382006-05-11 10:00:21 +0300219 ISCSI_TX_CONTROL ,
220 ISCSI_TX_SCSI_COMMAND,
221 ISCSI_TX_DATAOUT
222};
223
Or Gerlitzf19624a2010-02-08 13:19:56 +0000224struct iser_tx_desc {
Or Gerlitz49cd5382006-05-11 10:00:21 +0300225 struct iser_hdr iser_header;
226 struct iscsi_hdr iscsi_header;
Or Gerlitz49cd5382006-05-11 10:00:21 +0300227 enum iser_desc_type type;
Or Gerlitzf19624a2010-02-08 13:19:56 +0000228 u64 dma_addr;
229 /* sg[0] points to iser/iscsi headers, sg[1] optionally points to either
230 of immediate data, unsolicited data-out or control (login,text) */
231 struct ib_sge tx_sg[2];
232 int num_sge;
Or Gerlitz49cd5382006-05-11 10:00:21 +0300233};
234
Or Gerlitzbcc60c32010-02-08 13:17:42 +0000235#define ISER_RX_PAD_SIZE (256 - (ISER_RX_PAYLOAD_SIZE + \
236 sizeof(u64) + sizeof(struct ib_sge)))
237struct iser_rx_desc {
238 struct iser_hdr iser_header;
239 struct iscsi_hdr iscsi_header;
240 char data[ISER_RECV_DATA_SEG_LEN];
241 u64 dma_addr;
242 struct ib_sge rx_sg;
243 char pad[ISER_RX_PAD_SIZE];
244} __attribute__((packed));
245
Alex Tabachnik5a33a662012-09-23 15:17:44 +0000246#define ISER_MAX_CQ 4
247
Or Gerlitz49cd5382006-05-11 10:00:21 +0300248struct iser_device {
249 struct ib_device *ib_device;
250 struct ib_pd *pd;
Alex Tabachnik5a33a662012-09-23 15:17:44 +0000251 struct ib_cq *rx_cq[ISER_MAX_CQ];
252 struct ib_cq *tx_cq[ISER_MAX_CQ];
Or Gerlitz49cd5382006-05-11 10:00:21 +0300253 struct ib_mr *mr;
Alex Tabachnik5a33a662012-09-23 15:17:44 +0000254 struct tasklet_struct cq_tasklet[ISER_MAX_CQ];
Or Gerlitz2110f9b2010-05-05 17:30:10 +0300255 struct ib_event_handler event_handler;
Or Gerlitz49cd5382006-05-11 10:00:21 +0300256 struct list_head ig_list; /* entry in ig devices list */
257 int refcount;
Alex Tabachnik5a33a662012-09-23 15:17:44 +0000258 int cq_active_qps[ISER_MAX_CQ];
259 int cqs_used;
260 struct iser_cq_desc *cq_desc;
Or Gerlitz49cd5382006-05-11 10:00:21 +0300261};
262
263struct iser_conn {
264 struct iscsi_iser_conn *iser_conn; /* iser conn for upcalls */
Mike Christie412eeaf2008-05-21 15:54:14 -0500265 struct iscsi_endpoint *ep;
Or Gerlitz49cd5382006-05-11 10:00:21 +0300266 enum iser_ib_conn_state state; /* rdma connection state */
Mike Christieb40977d2008-05-21 15:54:03 -0500267 atomic_t refcount;
Or Gerlitz49cd5382006-05-11 10:00:21 +0300268 spinlock_t lock; /* used for state changes */
269 struct iser_device *device; /* device context */
270 struct rdma_cm_id *cma_id; /* CMA ID */
271 struct ib_qp *qp; /* QP */
272 struct ib_fmr_pool *fmr_pool; /* pool of IB FMRs */
Or Gerlitz49cd5382006-05-11 10:00:21 +0300273 wait_queue_head_t wait; /* waitq for conn/disconn */
Or Gerlitz704315f2010-02-08 13:18:39 +0000274 int post_recv_buf_count; /* posted rx count */
Or Gerlitz49cd5382006-05-11 10:00:21 +0300275 atomic_t post_send_buf_count; /* posted tx count */
Or Gerlitz49cd5382006-05-11 10:00:21 +0300276 char name[ISER_OBJECT_NAME_SIZE];
277 struct iser_page_vec *page_vec; /* represents SG to fmr maps*
278 * maps serialized as tx is*/
279 struct list_head conn_list; /* entry in ig conn list */
Or Gerlitzbcc60c32010-02-08 13:17:42 +0000280
281 char *login_buf;
Or Gerlitz2c4ce602011-11-04 00:19:46 +0200282 char *login_req_buf, *login_resp_buf;
283 u64 login_req_dma, login_resp_dma;
Or Gerlitzbcc60c32010-02-08 13:17:42 +0000284 unsigned int rx_desc_head;
285 struct iser_rx_desc *rx_descs;
286 struct ib_recv_wr rx_wr[ISER_MIN_POSTED_RX];
Or Gerlitz49cd5382006-05-11 10:00:21 +0300287};
288
289struct iscsi_iser_conn {
290 struct iscsi_conn *iscsi_conn;/* ptr to iscsi conn */
291 struct iser_conn *ib_conn; /* iSER IB conn */
Or Gerlitz49cd5382006-05-11 10:00:21 +0300292};
293
Mike Christie2261ec32008-05-21 15:54:11 -0500294struct iscsi_iser_task {
Or Gerlitzf19624a2010-02-08 13:19:56 +0000295 struct iser_tx_desc desc;
Or Gerlitz49cd5382006-05-11 10:00:21 +0300296 struct iscsi_iser_conn *iser_conn;
Or Gerlitz49cd5382006-05-11 10:00:21 +0300297 enum iser_task_status status;
298 int command_sent; /* set if command sent */
299 int dir[ISER_DIRS_NUM]; /* set if dir use*/
300 struct iser_regd_buf rdma_regd[ISER_DIRS_NUM];/* regd rdma buf */
301 struct iser_data_buf data[ISER_DIRS_NUM]; /* orig. data des*/
302 struct iser_data_buf data_copy[ISER_DIRS_NUM];/* contig. copy */
303};
304
305struct iser_page_vec {
306 u64 *pages;
307 int length;
308 int offset;
309 int data_size;
310};
311
Alex Tabachnik5a33a662012-09-23 15:17:44 +0000312struct iser_cq_desc {
313 struct iser_device *device;
314 int cq_index;
315};
316
Or Gerlitz49cd5382006-05-11 10:00:21 +0300317struct iser_global {
318 struct mutex device_list_mutex;/* */
319 struct list_head device_list; /* all iSER devices */
320 struct mutex connlist_mutex;
321 struct list_head connlist; /* all iSER IB connections */
322
Roland Dreiere54f8182006-11-29 15:33:07 -0800323 struct kmem_cache *desc_cache;
Or Gerlitz49cd5382006-05-11 10:00:21 +0300324};
325
326extern struct iser_global ig;
327extern int iser_debug_level;
328
329/* allocate connection resources needed for rdma functionality */
330int iser_conn_set_full_featured_mode(struct iscsi_conn *conn);
331
Mike Christie2747fdb2008-05-21 15:54:08 -0500332int iser_send_control(struct iscsi_conn *conn,
Mike Christie2261ec32008-05-21 15:54:11 -0500333 struct iscsi_task *task);
Or Gerlitz49cd5382006-05-11 10:00:21 +0300334
Mike Christie2747fdb2008-05-21 15:54:08 -0500335int iser_send_command(struct iscsi_conn *conn,
Mike Christie2261ec32008-05-21 15:54:11 -0500336 struct iscsi_task *task);
Or Gerlitz49cd5382006-05-11 10:00:21 +0300337
Mike Christie2747fdb2008-05-21 15:54:08 -0500338int iser_send_data_out(struct iscsi_conn *conn,
Mike Christie2261ec32008-05-21 15:54:11 -0500339 struct iscsi_task *task,
Mike Christie2747fdb2008-05-21 15:54:08 -0500340 struct iscsi_data *hdr);
Or Gerlitz49cd5382006-05-11 10:00:21 +0300341
342void iscsi_iser_recv(struct iscsi_conn *conn,
343 struct iscsi_hdr *hdr,
344 char *rx_data,
345 int rx_data_len);
346
Mike Christie412eeaf2008-05-21 15:54:14 -0500347void iser_conn_init(struct iser_conn *ib_conn);
Or Gerlitz49cd5382006-05-11 10:00:21 +0300348
Mike Christieb40977d2008-05-21 15:54:03 -0500349void iser_conn_get(struct iser_conn *ib_conn);
350
Or Gerlitz39ff05d2010-05-05 17:31:44 +0300351int iser_conn_put(struct iser_conn *ib_conn, int destroy_cma_id_allowed);
Or Gerlitz49cd5382006-05-11 10:00:21 +0300352
353void iser_conn_terminate(struct iser_conn *ib_conn);
354
Or Gerlitzbcc60c32010-02-08 13:17:42 +0000355void iser_rcv_completion(struct iser_rx_desc *desc,
356 unsigned long dto_xfer_len,
357 struct iser_conn *ib_conn);
Or Gerlitz49cd5382006-05-11 10:00:21 +0300358
Or Gerlitzf19624a2010-02-08 13:19:56 +0000359void iser_snd_completion(struct iser_tx_desc *desc, struct iser_conn *ib_conn);
Or Gerlitz49cd5382006-05-11 10:00:21 +0300360
Mike Christie2261ec32008-05-21 15:54:11 -0500361void iser_task_rdma_init(struct iscsi_iser_task *task);
Or Gerlitz49cd5382006-05-11 10:00:21 +0300362
Mike Christie2261ec32008-05-21 15:54:11 -0500363void iser_task_rdma_finalize(struct iscsi_iser_task *task);
Or Gerlitz49cd5382006-05-11 10:00:21 +0300364
Or Gerlitzbcc60c32010-02-08 13:17:42 +0000365void iser_free_rx_descriptors(struct iser_conn *ib_conn);
366
Mike Christie2261ec32008-05-21 15:54:11 -0500367void iser_finalize_rdma_unaligned_sg(struct iscsi_iser_task *task,
Or Gerlitz49cd5382006-05-11 10:00:21 +0300368 enum iser_data_dir cmd_dir);
369
Mike Christie2261ec32008-05-21 15:54:11 -0500370int iser_reg_rdma_mem(struct iscsi_iser_task *task,
Or Gerlitz49cd5382006-05-11 10:00:21 +0300371 enum iser_data_dir cmd_dir);
372
373int iser_connect(struct iser_conn *ib_conn,
374 struct sockaddr_in *src_addr,
375 struct sockaddr_in *dst_addr,
376 int non_blocking);
377
378int iser_reg_page_vec(struct iser_conn *ib_conn,
379 struct iser_page_vec *page_vec,
380 struct iser_mem_reg *mem_reg);
381
382void iser_unreg_mem(struct iser_mem_reg *mem_reg);
383
Or Gerlitzbcc60c32010-02-08 13:17:42 +0000384int iser_post_recvl(struct iser_conn *ib_conn);
385int iser_post_recvm(struct iser_conn *ib_conn, int count);
Or Gerlitzf19624a2010-02-08 13:19:56 +0000386int iser_post_send(struct iser_conn *ib_conn, struct iser_tx_desc *tx_desc);
Or Gerlitz49cd5382006-05-11 10:00:21 +0300387
Mike Christie2261ec32008-05-21 15:54:11 -0500388int iser_dma_map_task_data(struct iscsi_iser_task *iser_task,
Erez Zilber74a20782006-09-27 16:43:06 +0300389 struct iser_data_buf *data,
390 enum iser_data_dir iser_dir,
391 enum dma_data_direction dma_dir);
392
Mike Christie2261ec32008-05-21 15:54:11 -0500393void iser_dma_unmap_task_data(struct iscsi_iser_task *iser_task);
Or Gerlitzf19624a2010-02-08 13:19:56 +0000394int iser_initialize_task_headers(struct iscsi_task *task,
395 struct iser_tx_desc *tx_desc);
Or Gerlitz89e984e2012-03-05 18:21:44 +0200396int iser_alloc_rx_descriptors(struct iser_conn *ib_conn);
Or Gerlitz49cd5382006-05-11 10:00:21 +0300397#endif