2 * iSER transport for the Open iSCSI Initiator & iSER transport internals
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
9 * Copyright (c) 2004, 2005, 2006 Voltaire, Inc. All rights reserved.
10 * Copyright (c) 2005, 2006 Cisco Systems. All rights reserved.
11 * Copyright (c) 2013-2014 Mellanox Technologies. All rights reserved.
13 * This software is available to you under a choice of one of two
14 * licenses. You may choose to be licensed under the terms of the GNU
15 * General Public License (GPL) Version 2, available from the file
16 * COPYING in the main directory of this source tree, or the
17 * OpenIB.org BSD license below:
19 * Redistribution and use in source and binary forms, with or
20 * without modification, are permitted provided that the following
23 * - Redistributions of source code must retain the above
24 * copyright notice, this list of conditions and the following
27 * - Redistributions in binary form must reproduce the above
28 * copyright notice, this list of conditions and the following
29 * disclaimer in the documentation and/or other materials
30 * provided with the distribution.
32 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
33 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
34 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
35 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
36 * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
37 * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
38 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
41 #ifndef __ISCSI_ISER_H__
42 #define __ISCSI_ISER_H__
44 #include <linux/types.h>
45 #include <linux/net.h>
46 #include <linux/printk.h>
47 #include <scsi/libiscsi.h>
48 #include <scsi/scsi_transport_iscsi.h>
49 #include <scsi/scsi_cmnd.h>
50 #include <scsi/scsi_device.h>
52 #include <linux/interrupt.h>
53 #include <linux/wait.h>
54 #include <linux/sched.h>
55 #include <linux/list.h>
56 #include <linux/slab.h>
57 #include <linux/dma-mapping.h>
58 #include <linux/mutex.h>
59 #include <linux/mempool.h>
60 #include <linux/uio.h>
62 #include <linux/socket.h>
64 #include <linux/in6.h>
66 #include <rdma/ib_verbs.h>
67 #include <rdma/ib_fmr_pool.h>
68 #include <rdma/rdma_cm.h>
70 #define DRV_NAME "iser"
71 #define PFX DRV_NAME ": "
74 #define iser_dbg(fmt, arg...) \
76 if (unlikely(iser_debug_level > 2)) \
77 printk(KERN_DEBUG PFX "%s: " fmt,\
81 #define iser_warn(fmt, arg...) \
83 if (unlikely(iser_debug_level > 0)) \
84 pr_warn(PFX "%s: " fmt, \
88 #define iser_info(fmt, arg...) \
90 if (unlikely(iser_debug_level > 1)) \
91 pr_info(PFX "%s: " fmt, \
95 #define iser_err(fmt, arg...) \
96 pr_err(PFX "%s: " fmt, __func__ , ## arg)
99 #define SIZE_4K (1ULL << SHIFT_4K)
100 #define MASK_4K (~(SIZE_4K-1))
101 /* support up to 512KB in one RDMA */
102 #define ISCSI_ISER_SG_TABLESIZE (0x80000 >> SHIFT_4K)
103 #define ISER_DEF_XMIT_CMDS_DEFAULT 512
104 #if ISCSI_DEF_XMIT_CMDS_MAX > ISER_DEF_XMIT_CMDS_DEFAULT
105 #define ISER_DEF_XMIT_CMDS_MAX ISCSI_DEF_XMIT_CMDS_MAX
107 #define ISER_DEF_XMIT_CMDS_MAX ISER_DEF_XMIT_CMDS_DEFAULT
109 #define ISER_DEF_CMD_PER_LUN ISER_DEF_XMIT_CMDS_MAX
112 /* Maximal bounds on received asynchronous PDUs */
113 #define ISER_MAX_RX_MISC_PDUS 4 /* NOOP_IN(2) , ASYNC_EVENT(2) */
115 #define ISER_MAX_TX_MISC_PDUS 6 /* NOOP_OUT(2), TEXT(1), *
116 * SCSI_TMFUNC(2), LOGOUT(1) */
118 #define ISER_QP_MAX_RECV_DTOS (ISER_DEF_XMIT_CMDS_MAX)
120 #define ISER_MIN_POSTED_RX (ISER_DEF_XMIT_CMDS_MAX >> 2)
122 /* the max TX (send) WR supported by the iSER QP is defined by *
123 * max_send_wr = T * (1 + D) + C ; D is how many inflight dataouts we expect *
124 * to have at max for SCSI command. The tx posting & completion handling code *
125 * supports -EAGAIN scheme where tx is suspended till the QP has room for more *
126 * send WR. D=8 comes from 64K/8K */
128 #define ISER_INFLIGHT_DATAOUTS 8
130 #define ISER_QP_MAX_REQ_DTOS (ISER_DEF_XMIT_CMDS_MAX * \
131 (1 + ISER_INFLIGHT_DATAOUTS) + \
132 ISER_MAX_TX_MISC_PDUS + \
133 ISER_MAX_RX_MISC_PDUS)
135 /* Max registration work requests per command */
136 #define ISER_MAX_REG_WR_PER_CMD 5
138 /* For Signature we don't support DATAOUTs so no need to make room for them */
139 #define ISER_QP_SIG_MAX_REQ_DTOS (ISER_DEF_XMIT_CMDS_MAX * \
140 (1 + ISER_MAX_REG_WR_PER_CMD) + \
141 ISER_MAX_TX_MISC_PDUS + \
142 ISER_MAX_RX_MISC_PDUS)
144 #define ISER_GET_MAX_XMIT_CMDS(send_wr) ((send_wr \
145 - ISER_MAX_TX_MISC_PDUS \
146 - ISER_MAX_RX_MISC_PDUS) / \
147 (1 + ISER_INFLIGHT_DATAOUTS))
149 #define ISER_WC_BATCH_COUNT 16
150 #define ISER_SIGNAL_CMD_COUNT 32
152 #define ISER_VER 0x10
153 #define ISER_WSV 0x08
154 #define ISER_RSV 0x04
156 #define ISER_FASTREG_LI_WRID 0xffffffffffffffffULL
157 #define ISER_BEACON_WRID 0xfffffffffffffffeULL
160 * struct iser_hdr - iSER header
162 * @flags: flags support (zbva, remote_inv)
164 * @write_stag: write rkey
165 * @write_va: write virtual address
166 * @reaf_stag: read rkey
167 * @read_va: read virtual address
176 } __attribute__((packed
));
179 #define ISER_ZBVA_NOT_SUPPORTED 0x80
180 #define ISER_SEND_W_INV_NOT_SUPPORTED 0x40
187 /* Constant PDU lengths calculations */
188 #define ISER_HEADERS_LEN (sizeof(struct iser_hdr) + sizeof(struct iscsi_hdr))
190 #define ISER_RECV_DATA_SEG_LEN 128
191 #define ISER_RX_PAYLOAD_SIZE (ISER_HEADERS_LEN + ISER_RECV_DATA_SEG_LEN)
192 #define ISER_RX_LOGIN_SIZE (ISER_HEADERS_LEN + ISCSI_DEF_MAX_RECV_SEG_LEN)
194 /* Length of an object name string */
195 #define ISER_OBJECT_NAME_SIZE 64
197 enum iser_conn_state
{
198 ISER_CONN_INIT
, /* descriptor allocd, no conn */
199 ISER_CONN_PENDING
, /* in the process of being established */
200 ISER_CONN_UP
, /* up and running */
201 ISER_CONN_TERMINATING
, /* in the process of being terminated */
202 ISER_CONN_DOWN
, /* shut down */
206 enum iser_task_status
{
207 ISER_TASK_STATUS_INIT
= 0,
208 ISER_TASK_STATUS_STARTED
,
209 ISER_TASK_STATUS_COMPLETED
213 ISER_DIR_IN
= 0, /* to initiator */
214 ISER_DIR_OUT
, /* from initiator */
219 * struct iser_data_buf - iSER data buffer
221 * @sg: pointer to the sg list
222 * @size: num entries of this sg
223 * @data_len: total beffer byte len
224 * @dma_nents: returned by dma_map_sg
225 * @orig_sg: pointer to the original sg list (in case
227 * @orig_size: num entris of orig sg list
229 struct iser_data_buf
{
230 struct scatterlist
*sg
;
232 unsigned long data_len
;
233 unsigned int dma_nents
;
234 struct scatterlist
*orig_sg
;
235 unsigned int orig_size
;
238 /* fwd declarations */
240 struct iscsi_iser_task
;
241 struct iscsi_endpoint
;
244 * struct iser_mem_reg - iSER memory registration info
246 * @sge: memory region sg element
247 * @rkey: memory region remote key
248 * @mem_h: pointer to registration context (FMR/Fastreg)
250 struct iser_mem_reg
{
256 enum iser_desc_type
{
258 ISCSI_TX_SCSI_COMMAND
,
263 * struct iser_tx_desc - iSER TX descriptor (for send wr_id)
265 * @iser_header: iser header
266 * @iscsi_header: iscsi header
267 * @type: command/control/dataout
268 * @dam_addr: header buffer dma_address
269 * @tx_sg: sg[0] points to iser/iscsi headers
270 * sg[1] optionally points to either of immediate data
271 * unsolicited data-out or control
272 * @num_sge: number sges used on this TX task
274 struct iser_tx_desc
{
275 struct iser_hdr iser_header
;
276 struct iscsi_hdr iscsi_header
;
277 enum iser_desc_type type
;
279 struct ib_sge tx_sg
[2];
283 #define ISER_RX_PAD_SIZE (256 - (ISER_RX_PAYLOAD_SIZE + \
284 sizeof(u64) + sizeof(struct ib_sge)))
286 * struct iser_rx_desc - iSER RX descriptor (for recv wr_id)
288 * @iser_header: iser header
289 * @iscsi_header: iscsi header
290 * @data: received data segment
291 * @dma_addr: receive buffer dma address
292 * @rx_sg: ib_sge of receive buffer
293 * @pad: for sense data TODO: Modify to maximum sense length supported
295 struct iser_rx_desc
{
296 struct iser_hdr iser_header
;
297 struct iscsi_hdr iscsi_header
;
298 char data
[ISER_RECV_DATA_SEG_LEN
];
301 char pad
[ISER_RX_PAD_SIZE
];
302 } __attribute__((packed
));
306 struct iscsi_iser_task
;
309 * struct iser_comp - iSER completion context
311 * @device: pointer to device handle
312 * @cq: completion queue
313 * @wcs: work completion array
314 * @tasklet: Tasklet handle
315 * @active_qps: Number of active QPs attached
316 * to completion context
319 struct iser_device
*device
;
321 struct ib_wc wcs
[ISER_WC_BATCH_COUNT
];
322 struct tasklet_struct tasklet
;
327 * struct iser_device - iSER device handle
329 * @ib_device: RDMA device
330 * @pd: Protection Domain for this device
331 * @dev_attr: Device attributes container
332 * @mr: Global DMA memory region
333 * @event_handler: IB events handle routine
334 * @ig_list: entry in devices list
335 * @refcount: Reference counter, dominated by open iser connections
336 * @comps_used: Number of completion contexts used, Min between online
337 * cpus and device max completion vectors
338 * @comps: Dinamically allocated array of completion handlers
339 * Memory registration pool Function pointers (FMR or Fastreg):
340 * @iser_alloc_rdma_reg_res: Allocation of memory regions pool
341 * @iser_free_rdma_reg_res: Free of memory regions pool
342 * @iser_reg_rdma_mem: Memory registration routine
343 * @iser_unreg_rdma_mem: Memory deregistration routine
346 struct ib_device
*ib_device
;
348 struct ib_device_attr dev_attr
;
350 struct ib_event_handler event_handler
;
351 struct list_head ig_list
;
354 struct iser_comp
*comps
;
355 int (*iser_alloc_rdma_reg_res
)(struct ib_conn
*ib_conn
,
357 void (*iser_free_rdma_reg_res
)(struct ib_conn
*ib_conn
);
358 int (*iser_reg_rdma_mem
)(struct iscsi_iser_task
*iser_task
,
359 enum iser_data_dir cmd_dir
);
360 void (*iser_unreg_rdma_mem
)(struct iscsi_iser_task
*iser_task
,
361 enum iser_data_dir cmd_dir
);
364 #define ISER_CHECK_GUARD 0xc0
365 #define ISER_CHECK_REFTAG 0x0f
366 #define ISER_CHECK_APPTAG 0x30
368 enum iser_reg_indicator
{
369 ISER_DATA_KEY_VALID
= 1 << 0,
370 ISER_PROT_KEY_VALID
= 1 << 1,
371 ISER_SIG_KEY_VALID
= 1 << 2,
372 ISER_FASTREG_PROTECTED
= 1 << 3,
376 * struct iser_pi_context - Protection information context
378 * @prot_mr: protection memory region
379 * @prot_frpl: protection fastreg page list
380 * @sig_mr: signature feature enabled memory region
382 struct iser_pi_context
{
383 struct ib_mr
*prot_mr
;
384 struct ib_fast_reg_page_list
*prot_frpl
;
385 struct ib_mr
*sig_mr
;
389 * struct fast_reg_descriptor - Fast registration descriptor
391 * @list: entry in connection fastreg pool
392 * @data_mr: data memory region
393 * @data_frpl: data fastreg page list
394 * @pi_ctx: protection information context
395 * @reg_indicators: fast registration indicators
397 struct fast_reg_descriptor
{
398 struct list_head list
;
399 struct ib_mr
*data_mr
;
400 struct ib_fast_reg_page_list
*data_frpl
;
401 struct iser_pi_context
*pi_ctx
;
406 * struct ib_conn - Infiniband related objects
408 * @cma_id: rdma_cm connection maneger handle
409 * @qp: Connection Queue-pair
410 * @post_recv_buf_count: post receive counter
411 * @sig_count: send work request signal count
412 * @rx_wr: receive work request for batch posts
413 * @device: reference to iser device
414 * @comp: iser completion context
415 * @pi_support: Indicate device T10-PI support
416 * @beacon: beacon send wr to signal all flush errors were drained
417 * @flush_comp: completes when all connection completions consumed
418 * @lock: protects fmr/fastreg pool
420 * @pool: FMR pool for fast registrations
421 * @page_vec: page vector to hold mapped commands pages
422 * used for registration
424 * @pool: Fast registration descriptors pool for fast
426 * @pool_size: Size of pool
429 struct rdma_cm_id
*cma_id
;
431 int post_recv_buf_count
;
433 struct ib_recv_wr rx_wr
[ISER_MIN_POSTED_RX
];
434 struct iser_device
*device
;
435 struct iser_comp
*comp
;
437 struct ib_send_wr beacon
;
438 struct completion flush_comp
;
442 struct ib_fmr_pool
*pool
;
443 struct iser_page_vec
*page_vec
;
446 struct list_head pool
;
453 * struct iser_conn - iSER connection context
455 * @ib_conn: connection RDMA resources
456 * @iscsi_conn: link to matching iscsi connection
457 * @ep: transport handle
458 * @state: connection logical state
459 * @qp_max_recv_dtos: maximum number of data outs, corresponds
460 * to max number of post recvs
461 * @qp_max_recv_dtos_mask: (qp_max_recv_dtos - 1)
462 * @min_posted_rx: (qp_max_recv_dtos >> 2)
463 * @max_cmds: maximum cmds allowed for this connection
464 * @name: connection peer portal
465 * @release_work: deffered work for release job
466 * @state_mutex: protects iser onnection state
467 * @stop_completion: conn_stop completion
468 * @ib_completion: RDMA cleanup completion
469 * @up_completion: connection establishment completed
470 * (state is ISER_CONN_UP)
471 * @conn_list: entry in ig conn list
472 * @login_buf: login data buffer (stores login parameters)
473 * @login_req_buf: login request buffer
474 * @login_req_dma: login request buffer dma address
475 * @login_resp_buf: login response buffer
476 * @login_resp_dma: login response buffer dma address
477 * @rx_desc_head: head of rx_descs cyclic buffer
478 * @rx_descs: rx buffers array (cyclic buffer)
479 * @num_rx_descs: number of rx descriptors
482 struct ib_conn ib_conn
;
483 struct iscsi_conn
*iscsi_conn
;
484 struct iscsi_endpoint
*ep
;
485 enum iser_conn_state state
;
486 unsigned qp_max_recv_dtos
;
487 unsigned qp_max_recv_dtos_mask
;
488 unsigned min_posted_rx
;
490 char name
[ISER_OBJECT_NAME_SIZE
];
491 struct work_struct release_work
;
492 struct mutex state_mutex
;
493 struct completion stop_completion
;
494 struct completion ib_completion
;
495 struct completion up_completion
;
496 struct list_head conn_list
;
499 char *login_req_buf
, *login_resp_buf
;
500 u64 login_req_dma
, login_resp_dma
;
501 unsigned int rx_desc_head
;
502 struct iser_rx_desc
*rx_descs
;
507 * struct iscsi_iser_task - iser task context
509 * @desc: TX descriptor
510 * @iser_conn: link to iser connection
511 * @status: current task status
512 * @sc: link to scsi command
513 * @command_sent: indicate if command was sent
514 * @dir: iser data direction
515 * @rdma_reg: task rdma registration desc
516 * @data: iser data buffer desc
517 * @prot: iser protection buffer desc
519 struct iscsi_iser_task
{
520 struct iser_tx_desc desc
;
521 struct iser_conn
*iser_conn
;
522 enum iser_task_status status
;
523 struct scsi_cmnd
*sc
;
525 int dir
[ISER_DIRS_NUM
];
526 struct iser_mem_reg rdma_reg
[ISER_DIRS_NUM
];
527 struct iser_data_buf data
[ISER_DIRS_NUM
];
528 struct iser_data_buf prot
[ISER_DIRS_NUM
];
531 struct iser_page_vec
{
539 * struct iser_global: iSER global context
541 * @device_list_mutex: protects device_list
542 * @device_list: iser devices global list
543 * @connlist_mutex: protects connlist
544 * @connlist: iser connections global list
545 * @desc_cache: kmem cache for tx dataout
548 struct mutex device_list_mutex
;
549 struct list_head device_list
;
550 struct mutex connlist_mutex
;
551 struct list_head connlist
;
552 struct kmem_cache
*desc_cache
;
555 extern struct iser_global ig
;
556 extern int iser_debug_level
;
557 extern bool iser_pi_enable
;
558 extern int iser_pi_guard
;
560 int iser_send_control(struct iscsi_conn
*conn
,
561 struct iscsi_task
*task
);
563 int iser_send_command(struct iscsi_conn
*conn
,
564 struct iscsi_task
*task
);
566 int iser_send_data_out(struct iscsi_conn
*conn
,
567 struct iscsi_task
*task
,
568 struct iscsi_data
*hdr
);
570 void iscsi_iser_recv(struct iscsi_conn
*conn
,
571 struct iscsi_hdr
*hdr
,
575 void iser_conn_init(struct iser_conn
*iser_conn
);
577 void iser_conn_release(struct iser_conn
*iser_conn
);
579 int iser_conn_terminate(struct iser_conn
*iser_conn
);
581 void iser_release_work(struct work_struct
*work
);
583 void iser_rcv_completion(struct iser_rx_desc
*desc
,
584 unsigned long dto_xfer_len
,
585 struct ib_conn
*ib_conn
);
587 void iser_snd_completion(struct iser_tx_desc
*desc
,
588 struct ib_conn
*ib_conn
);
590 void iser_task_rdma_init(struct iscsi_iser_task
*task
);
592 void iser_task_rdma_finalize(struct iscsi_iser_task
*task
);
594 void iser_free_rx_descriptors(struct iser_conn
*iser_conn
);
596 void iser_finalize_rdma_unaligned_sg(struct iscsi_iser_task
*iser_task
,
597 struct iser_data_buf
*mem
,
598 enum iser_data_dir cmd_dir
);
600 int iser_reg_rdma_mem_fmr(struct iscsi_iser_task
*task
,
601 enum iser_data_dir cmd_dir
);
602 int iser_reg_rdma_mem_fastreg(struct iscsi_iser_task
*task
,
603 enum iser_data_dir cmd_dir
);
605 int iser_connect(struct iser_conn
*iser_conn
,
606 struct sockaddr
*src_addr
,
607 struct sockaddr
*dst_addr
,
610 void iser_unreg_mem_fmr(struct iscsi_iser_task
*iser_task
,
611 enum iser_data_dir cmd_dir
);
612 void iser_unreg_mem_fastreg(struct iscsi_iser_task
*iser_task
,
613 enum iser_data_dir cmd_dir
);
615 int iser_post_recvl(struct iser_conn
*iser_conn
);
616 int iser_post_recvm(struct iser_conn
*iser_conn
, int count
);
617 int iser_post_send(struct ib_conn
*ib_conn
, struct iser_tx_desc
*tx_desc
,
620 int iser_dma_map_task_data(struct iscsi_iser_task
*iser_task
,
621 struct iser_data_buf
*data
,
622 enum iser_data_dir iser_dir
,
623 enum dma_data_direction dma_dir
);
625 void iser_dma_unmap_task_data(struct iscsi_iser_task
*iser_task
,
626 struct iser_data_buf
*data
,
627 enum dma_data_direction dir
);
629 int iser_initialize_task_headers(struct iscsi_task
*task
,
630 struct iser_tx_desc
*tx_desc
);
631 int iser_alloc_rx_descriptors(struct iser_conn
*iser_conn
,
632 struct iscsi_session
*session
);
633 int iser_create_fmr_pool(struct ib_conn
*ib_conn
, unsigned cmds_max
);
634 void iser_free_fmr_pool(struct ib_conn
*ib_conn
);
635 int iser_create_fastreg_pool(struct ib_conn
*ib_conn
, unsigned cmds_max
);
636 void iser_free_fastreg_pool(struct ib_conn
*ib_conn
);
637 u8
iser_check_task_pi_status(struct iscsi_iser_task
*iser_task
,
638 enum iser_data_dir cmd_dir
, sector_t
*sector
);
639 struct fast_reg_descriptor
*
640 iser_reg_desc_get(struct ib_conn
*ib_conn
);
642 iser_reg_desc_put(struct ib_conn
*ib_conn
,
643 struct fast_reg_descriptor
*desc
);