xref: /spdk/lib/nvmf/nvmf_internal.h (revision 73f79a5c56823bb53e2891c7e16a961b3c192fbc)
1 /*-
2  *   BSD LICENSE
3  *
4  *   Copyright (c) Intel Corporation.
5  *   All rights reserved.
6  *
7  *   Redistribution and use in source and binary forms, with or without
8  *   modification, are permitted provided that the following conditions
9  *   are met:
10  *
11  *     * Redistributions of source code must retain the above copyright
12  *       notice, this list of conditions and the following disclaimer.
13  *     * Redistributions in binary form must reproduce the above copyright
14  *       notice, this list of conditions and the following disclaimer in
15  *       the documentation and/or other materials provided with the
16  *       distribution.
17  *     * Neither the name of Intel Corporation nor the names of its
18  *       contributors may be used to endorse or promote products derived
19  *       from this software without specific prior written permission.
20  *
21  *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
22  *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
23  *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
24  *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
25  *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
26  *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
27  *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
28  *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
29  *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
30  *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
31  *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
32  */
33 
34 #ifndef __NVMF_INTERNAL_H__
35 #define __NVMF_INTERNAL_H__
36 
37 #include "spdk/stdinc.h"
38 
39 #include "spdk/likely.h"
40 #include "spdk/nvmf.h"
41 #include "spdk/nvmf_spec.h"
42 #include "spdk/assert.h"
43 #include "spdk/bdev.h"
44 #include "spdk/queue.h"
45 #include "spdk/util.h"
46 #include "spdk/thread.h"
47 
48 #define SPDK_NVMF_MAX_SGL_ENTRIES	16
49 
50 /* AIO backend requires block size aligned data buffers,
51  * extra 4KiB aligned data buffer should work for most devices.
52  */
53 #define SHIFT_4KB			12u
54 #define NVMF_DATA_BUFFER_ALIGNMENT	(1u << SHIFT_4KB)
55 #define NVMF_DATA_BUFFER_MASK		(NVMF_DATA_BUFFER_ALIGNMENT - 1LL)
56 
57 enum spdk_nvmf_subsystem_state {
58 	SPDK_NVMF_SUBSYSTEM_INACTIVE = 0,
59 	SPDK_NVMF_SUBSYSTEM_ACTIVATING,
60 	SPDK_NVMF_SUBSYSTEM_ACTIVE,
61 	SPDK_NVMF_SUBSYSTEM_PAUSING,
62 	SPDK_NVMF_SUBSYSTEM_PAUSED,
63 	SPDK_NVMF_SUBSYSTEM_RESUMING,
64 	SPDK_NVMF_SUBSYSTEM_DEACTIVATING,
65 };
66 
67 enum spdk_nvmf_qpair_state {
68 	SPDK_NVMF_QPAIR_UNINITIALIZED = 0,
69 	SPDK_NVMF_QPAIR_ACTIVATING,
70 	SPDK_NVMF_QPAIR_ACTIVE,
71 	SPDK_NVMF_QPAIR_DEACTIVATING,
72 	SPDK_NVMF_QPAIR_ERROR,
73 };
74 
75 typedef void (*spdk_nvmf_state_change_done)(void *cb_arg, int status);
76 
77 struct spdk_nvmf_tgt {
78 	uint64_t				discovery_genctr;
79 
80 	uint32_t				max_subsystems;
81 
82 	/* Array of subsystem pointers of size max_subsystems indexed by sid */
83 	struct spdk_nvmf_subsystem		**subsystems;
84 
85 	struct spdk_nvmf_discovery_log_page	*discovery_log_page;
86 	size_t					discovery_log_page_size;
87 	TAILQ_HEAD(, spdk_nvmf_transport)	transports;
88 
89 	spdk_nvmf_tgt_destroy_done_fn		*destroy_cb_fn;
90 	void					*destroy_cb_arg;
91 };
92 
93 struct spdk_nvmf_host {
94 	char				*nqn;
95 	TAILQ_ENTRY(spdk_nvmf_host)	link;
96 };
97 
98 struct spdk_nvmf_listener {
99 	struct spdk_nvme_transport_id	trid;
100 	struct spdk_nvmf_transport	*transport;
101 	TAILQ_ENTRY(spdk_nvmf_listener)	link;
102 };
103 
104 struct spdk_nvmf_transport_pg_cache_buf {
105 	STAILQ_ENTRY(spdk_nvmf_transport_pg_cache_buf) link;
106 };
107 
108 struct spdk_nvmf_transport_poll_group {
109 	struct spdk_nvmf_transport					*transport;
110 	STAILQ_HEAD(, spdk_nvmf_transport_pg_cache_buf)			buf_cache;
111 	uint32_t							buf_cache_count;
112 	uint32_t							buf_cache_size;
113 	TAILQ_ENTRY(spdk_nvmf_transport_poll_group)			link;
114 };
115 
116 struct spdk_nvmf_subsystem_poll_group {
117 	/* Array of channels for each namespace indexed by nsid - 1 */
118 	struct spdk_io_channel	**channels;
119 	uint32_t		num_channels;
120 
121 	enum spdk_nvmf_subsystem_state state;
122 
123 	TAILQ_HEAD(, spdk_nvmf_request)	queued;
124 };
125 
126 struct spdk_nvmf_poll_group {
127 	struct spdk_thread				*thread;
128 	struct spdk_poller				*poller;
129 
130 	TAILQ_HEAD(, spdk_nvmf_transport_poll_group)	tgroups;
131 
132 	/* Array of poll groups indexed by subsystem id (sid) */
133 	struct spdk_nvmf_subsystem_poll_group		*sgroups;
134 	uint32_t					num_sgroups;
135 
136 	/* All of the queue pairs that belong to this poll group */
137 	TAILQ_HEAD(, spdk_nvmf_qpair)			qpairs;
138 };
139 
140 typedef enum _spdk_nvmf_request_exec_status {
141 	SPDK_NVMF_REQUEST_EXEC_STATUS_COMPLETE,
142 	SPDK_NVMF_REQUEST_EXEC_STATUS_ASYNCHRONOUS,
143 } spdk_nvmf_request_exec_status;
144 
145 union nvmf_h2c_msg {
146 	struct spdk_nvmf_capsule_cmd			nvmf_cmd;
147 	struct spdk_nvme_cmd				nvme_cmd;
148 	struct spdk_nvmf_fabric_prop_set_cmd		prop_set_cmd;
149 	struct spdk_nvmf_fabric_prop_get_cmd		prop_get_cmd;
150 	struct spdk_nvmf_fabric_connect_cmd		connect_cmd;
151 };
152 SPDK_STATIC_ASSERT(sizeof(union nvmf_h2c_msg) == 64, "Incorrect size");
153 
154 union nvmf_c2h_msg {
155 	struct spdk_nvme_cpl				nvme_cpl;
156 	struct spdk_nvmf_fabric_prop_get_rsp		prop_get_rsp;
157 	struct spdk_nvmf_fabric_connect_rsp		connect_rsp;
158 };
159 SPDK_STATIC_ASSERT(sizeof(union nvmf_c2h_msg) == 16, "Incorrect size");
160 
161 struct spdk_nvmf_request {
162 	struct spdk_nvmf_qpair		*qpair;
163 	uint32_t			length;
164 	enum spdk_nvme_data_transfer	xfer;
165 	void				*data;
166 	union nvmf_h2c_msg		*cmd;
167 	union nvmf_c2h_msg		*rsp;
168 	struct iovec			iov[SPDK_NVMF_MAX_SGL_ENTRIES * 2];
169 	uint32_t			iovcnt;
170 	struct spdk_bdev_io_wait_entry	bdev_io_wait;
171 
172 	TAILQ_ENTRY(spdk_nvmf_request)	link;
173 };
174 
175 struct spdk_nvmf_ns {
176 	uint32_t nsid;
177 	struct spdk_nvmf_subsystem *subsystem;
178 	struct spdk_bdev *bdev;
179 	struct spdk_bdev_desc *desc;
180 	struct spdk_nvmf_ns_opts opts;
181 };
182 
183 struct spdk_nvmf_qpair {
184 	enum spdk_nvmf_qpair_state		state;
185 	spdk_nvmf_state_change_done		state_cb;
186 	void					*state_cb_arg;
187 
188 	struct spdk_nvmf_transport		*transport;
189 	struct spdk_nvmf_ctrlr			*ctrlr;
190 	struct spdk_nvmf_poll_group		*group;
191 
192 	uint16_t				qid;
193 	uint16_t				sq_head;
194 	uint16_t				sq_head_max;
195 
196 	TAILQ_HEAD(, spdk_nvmf_request)		outstanding;
197 	TAILQ_ENTRY(spdk_nvmf_qpair)		link;
198 };
199 
200 struct spdk_nvmf_ctrlr_feat {
201 	union spdk_nvme_feat_arbitration arbitration;
202 	union spdk_nvme_feat_power_management power_management;
203 	union spdk_nvme_feat_error_recovery error_recovery;
204 	union spdk_nvme_feat_volatile_write_cache volatile_write_cache;
205 	union spdk_nvme_feat_number_of_queues number_of_queues;
206 	union spdk_nvme_feat_write_atomicity write_atomicity;
207 	union spdk_nvme_feat_async_event_configuration async_event_configuration;
208 	union spdk_nvme_feat_keep_alive_timer keep_alive_timer;
209 };
210 
211 /*
212  * This structure represents an NVMe-oF controller,
213  * which is like a "session" in networking terms.
214  */
215 struct spdk_nvmf_ctrlr {
216 	uint16_t			cntlid;
217 	struct spdk_nvmf_subsystem	*subsys;
218 
219 	struct {
220 		union spdk_nvme_cap_register	cap;
221 		union spdk_nvme_vs_register	vs;
222 		union spdk_nvme_cc_register	cc;
223 		union spdk_nvme_csts_register	csts;
224 	} vcprop; /* virtual controller properties */
225 
226 	struct spdk_nvmf_ctrlr_feat feat;
227 
228 	struct spdk_nvmf_qpair	*admin_qpair;
229 	struct spdk_thread	*thread;
230 	struct spdk_bit_array	*qpair_mask;
231 
232 	struct spdk_nvmf_request *aer_req;
233 	union spdk_nvme_async_event_completion notice_event;
234 	struct spdk_uuid  hostid;
235 
236 	uint16_t changed_ns_list_count;
237 	struct spdk_nvme_ns_list changed_ns_list;
238 
239 	/* Time to trigger keep-alive--poller_time = now_tick + period */
240 	uint64_t last_keep_alive_tick;
241 	struct spdk_poller			*keep_alive_poller;
242 
243 	TAILQ_ENTRY(spdk_nvmf_ctrlr)		link;
244 };
245 
246 struct spdk_nvmf_subsystem {
247 	struct spdk_thread		*thread;
248 	uint32_t			id;
249 	enum spdk_nvmf_subsystem_state	state;
250 
251 	char subnqn[SPDK_NVMF_NQN_MAX_LEN + 1];
252 	enum spdk_nvmf_subtype subtype;
253 	uint16_t next_cntlid;
254 	bool allow_any_host;
255 
256 	struct spdk_nvmf_tgt			*tgt;
257 
258 	char sn[SPDK_NVME_CTRLR_SN_LEN + 1];
259 
260 	/* Array of pointers to namespaces of size max_nsid indexed by nsid - 1 */
261 	struct spdk_nvmf_ns			**ns;
262 	uint32_t				max_nsid;
263 	/* This is the maximum allowed nsid to a subsystem */
264 	uint32_t				max_allowed_nsid;
265 
266 	TAILQ_HEAD(, spdk_nvmf_ctrlr)		ctrlrs;
267 
268 	TAILQ_HEAD(, spdk_nvmf_host)		hosts;
269 
270 	TAILQ_HEAD(, spdk_nvmf_listener)	listeners;
271 
272 	TAILQ_ENTRY(spdk_nvmf_subsystem)	entries;
273 };
274 
275 typedef void(*spdk_nvmf_poll_group_mod_done)(void *cb_arg, int status);
276 
277 struct spdk_nvmf_transport *spdk_nvmf_tgt_get_transport(struct spdk_nvmf_tgt *tgt,
278 		enum spdk_nvme_transport_type);
279 
280 int spdk_nvmf_poll_group_add_transport(struct spdk_nvmf_poll_group *group,
281 				       struct spdk_nvmf_transport *transport);
282 int spdk_nvmf_poll_group_update_subsystem(struct spdk_nvmf_poll_group *group,
283 		struct spdk_nvmf_subsystem *subsystem);
284 int spdk_nvmf_poll_group_add_subsystem(struct spdk_nvmf_poll_group *group,
285 				       struct spdk_nvmf_subsystem *subsystem,
286 				       spdk_nvmf_poll_group_mod_done cb_fn, void *cb_arg);
287 void spdk_nvmf_poll_group_remove_subsystem(struct spdk_nvmf_poll_group *group,
288 		struct spdk_nvmf_subsystem *subsystem, spdk_nvmf_poll_group_mod_done cb_fn, void *cb_arg);
289 void spdk_nvmf_poll_group_pause_subsystem(struct spdk_nvmf_poll_group *group,
290 		struct spdk_nvmf_subsystem *subsystem, spdk_nvmf_poll_group_mod_done cb_fn, void *cb_arg);
291 void spdk_nvmf_poll_group_resume_subsystem(struct spdk_nvmf_poll_group *group,
292 		struct spdk_nvmf_subsystem *subsystem, spdk_nvmf_poll_group_mod_done cb_fn, void *cb_arg);
293 void spdk_nvmf_request_exec(struct spdk_nvmf_request *req);
294 int spdk_nvmf_request_free(struct spdk_nvmf_request *req);
295 int spdk_nvmf_request_complete(struct spdk_nvmf_request *req);
296 
297 void spdk_nvmf_get_discovery_log_page(struct spdk_nvmf_tgt *tgt, struct iovec *iov,
298 				      uint32_t iovcnt, uint64_t offset, uint32_t length);
299 
300 void spdk_nvmf_ctrlr_destruct(struct spdk_nvmf_ctrlr *ctrlr);
301 int spdk_nvmf_ctrlr_process_fabrics_cmd(struct spdk_nvmf_request *req);
302 int spdk_nvmf_ctrlr_process_admin_cmd(struct spdk_nvmf_request *req);
303 int spdk_nvmf_ctrlr_process_io_cmd(struct spdk_nvmf_request *req);
304 bool spdk_nvmf_ctrlr_dsm_supported(struct spdk_nvmf_ctrlr *ctrlr);
305 bool spdk_nvmf_ctrlr_write_zeroes_supported(struct spdk_nvmf_ctrlr *ctrlr);
306 void spdk_nvmf_ctrlr_ns_changed(struct spdk_nvmf_ctrlr *ctrlr, uint32_t nsid);
307 
308 void spdk_nvmf_bdev_ctrlr_identify_ns(struct spdk_nvmf_ns *ns, struct spdk_nvme_ns_data *nsdata);
309 
310 int spdk_nvmf_subsystem_add_ctrlr(struct spdk_nvmf_subsystem *subsystem,
311 				  struct spdk_nvmf_ctrlr *ctrlr);
312 void spdk_nvmf_subsystem_remove_ctrlr(struct spdk_nvmf_subsystem *subsystem,
313 				      struct spdk_nvmf_ctrlr *ctrlr);
314 struct spdk_nvmf_ctrlr *spdk_nvmf_subsystem_get_ctrlr(struct spdk_nvmf_subsystem *subsystem,
315 		uint16_t cntlid);
316 int spdk_nvmf_ctrlr_async_event_ns_notice(struct spdk_nvmf_ctrlr *ctrlr);
317 
318 /*
319  * Abort aer is sent on a per controller basis and sends a completion for the aer to the host.
320  * This function should be called when attempting to recover in error paths when it is OK for
321  * the host to send a subsequent AER.
322  */
323 void spdk_nvmf_ctrlr_abort_aer(struct spdk_nvmf_ctrlr *ctrlr);
324 
325 /*
326  * Free aer simply frees the rdma resources for the aer without informing the host.
327  * This function should be called when deleting a qpair when one wants to make sure
328  * the qpair is completely empty before freeing the request. The reason we free the
329  * AER without sending a completion is to prevent the host from sending another AER.
330  */
331 void spdk_nvmf_qpair_free_aer(struct spdk_nvmf_qpair *qpair);
332 
333 static inline struct spdk_nvmf_ns *
334 _spdk_nvmf_subsystem_get_ns(struct spdk_nvmf_subsystem *subsystem, uint32_t nsid)
335 {
336 	/* NOTE: This implicitly also checks for 0, since 0 - 1 wraps around to UINT32_MAX. */
337 	if (spdk_unlikely(nsid - 1 >= subsystem->max_nsid)) {
338 		return NULL;
339 	}
340 
341 	return subsystem->ns[nsid - 1];
342 }
343 
344 static inline bool
345 spdk_nvmf_qpair_is_admin_queue(struct spdk_nvmf_qpair *qpair)
346 {
347 	return qpair->qid == 0;
348 }
349 
350 #endif /* __NVMF_INTERNAL_H__ */
351