1 /* SPDX-License-Identifier: GPL-2.0 */
3 * Copyright (c) 2015-2016 HGST, a Western Digital Company.
9 #include <linux/dma-mapping.h>
10 #include <linux/types.h>
11 #include <linux/device.h>
12 #include <linux/kref.h>
13 #include <linux/percpu-refcount.h>
14 #include <linux/list.h>
15 #include <linux/mutex.h>
16 #include <linux/uuid.h>
17 #include <linux/nvme.h>
18 #include <linux/configfs.h>
19 #include <linux/rcupdate.h>
20 #include <linux/blkdev.h>
21 #include <linux/radix-tree.h>
23 #define NVMET_ASYNC_EVENTS 4
24 #define NVMET_ERROR_LOG_SLOTS 128
25 #define NVMET_NO_ERROR_LOC ((u16)-1)
28 * Supported optional AENs:
30 #define NVMET_AEN_CFG_OPTIONAL \
31 (NVME_AEN_CFG_NS_ATTR | NVME_AEN_CFG_ANA_CHANGE)
32 #define NVMET_DISC_AEN_CFG_OPTIONAL \
33 (NVME_AEN_CFG_DISC_CHANGE)
36 * Plus mandatory SMART AENs (we'll never send them, but allow enabling them):
38 #define NVMET_AEN_CFG_ALL \
39 (NVME_SMART_CRIT_SPARE | NVME_SMART_CRIT_TEMPERATURE | \
40 NVME_SMART_CRIT_RELIABILITY | NVME_SMART_CRIT_MEDIA | \
41 NVME_SMART_CRIT_VOLATILE_MEMORY | NVMET_AEN_CFG_OPTIONAL)
43 /* Helper Macros when NVMe error is NVME_SC_CONNECT_INVALID_PARAM
44 * The 16 bit shift is to set IATTR bit to 1, which means offending
45 * offset starts in the data section of connect()
47 #define IPO_IATTR_CONNECT_DATA(x) \
48 (cpu_to_le32((1 << 16) | (offsetof(struct nvmf_connect_data, x))))
49 #define IPO_IATTR_CONNECT_SQE(x) \
50 (cpu_to_le32(offsetof(struct nvmf_connect_command, x)))
53 struct list_head dev_link
;
54 struct percpu_ref ref
;
55 struct block_device
*bdev
;
67 struct nvmet_subsys
*subsys
;
68 const char *device_path
;
70 struct config_group device_group
;
71 struct config_group group
;
73 struct completion disable_done
;
75 struct kmem_cache
*bvec_cache
;
78 struct pci_dev
*p2p_dev
;
81 static inline struct nvmet_ns
*to_nvmet_ns(struct config_item
*item
)
83 return container_of(to_config_group(item
), struct nvmet_ns
, group
);
86 static inline struct device
*nvmet_ns_dev(struct nvmet_ns
*ns
)
88 return ns
->bdev
? disk_to_dev(ns
->bdev
->bd_disk
) : NULL
;
97 struct nvmet_ctrl
*ctrl
;
98 struct percpu_ref ref
;
103 struct completion free_done
;
104 struct completion confirm_done
;
107 struct nvmet_ana_group
{
108 struct config_group group
;
109 struct nvmet_port
*port
;
113 static inline struct nvmet_ana_group
*to_ana_group(struct config_item
*item
)
115 return container_of(to_config_group(item
), struct nvmet_ana_group
,
120 * struct nvmet_port - Common structure to keep port
121 * information for the target.
122 * @entry: Entry into referrals or transport list.
123 * @disc_addr: Address information is stored in a format defined
124 * for a discovery log page entry.
125 * @group: ConfigFS group for this element's folder.
126 * @priv: Private data for the transport.
129 struct list_head entry
;
130 struct nvmf_disc_rsp_page_entry disc_addr
;
131 struct config_group group
;
132 struct config_group subsys_group
;
133 struct list_head subsystems
;
134 struct config_group referrals_group
;
135 struct list_head referrals
;
136 struct list_head global_entry
;
137 struct config_group ana_groups_group
;
138 struct nvmet_ana_group ana_default_group
;
139 enum nvme_ana_state
*ana_state
;
142 int inline_data_size
;
143 const struct nvmet_fabrics_ops
*tr_ops
;
146 static inline struct nvmet_port
*to_nvmet_port(struct config_item
*item
)
148 return container_of(to_config_group(item
), struct nvmet_port
,
152 static inline struct nvmet_port
*ana_groups_to_port(
153 struct config_item
*item
)
155 return container_of(to_config_group(item
), struct nvmet_port
,
160 struct nvmet_subsys
*subsys
;
161 struct nvmet_cq
**cqs
;
162 struct nvmet_sq
**sqs
;
175 struct nvmet_port
*port
;
178 unsigned long aen_masked
;
179 struct nvmet_req
*async_event_cmds
[NVMET_ASYNC_EVENTS
];
180 unsigned int nr_async_event_cmds
;
181 struct list_head async_events
;
182 struct work_struct async_event_work
;
184 struct list_head subsys_entry
;
186 struct delayed_work ka_work
;
187 struct work_struct fatal_err_work
;
189 const struct nvmet_fabrics_ops
*ops
;
191 __le32
*changed_ns_list
;
194 char subsysnqn
[NVMF_NQN_FIELD_LEN
];
195 char hostnqn
[NVMF_NQN_FIELD_LEN
];
197 struct device
*p2p_client
;
198 struct radix_tree_root p2p_ns_map
;
200 spinlock_t error_lock
;
202 struct nvme_error_slot slots
[NVMET_ERROR_LOG_SLOTS
];
205 struct nvmet_subsys
{
206 enum nvme_subsys_type type
;
211 struct list_head namespaces
;
212 unsigned int nr_namespaces
;
213 unsigned int max_nsid
;
215 struct list_head ctrls
;
217 struct list_head hosts
;
226 struct config_group group
;
228 struct config_group namespaces_group
;
229 struct config_group allowed_hosts_group
;
232 static inline struct nvmet_subsys
*to_subsys(struct config_item
*item
)
234 return container_of(to_config_group(item
), struct nvmet_subsys
, group
);
237 static inline struct nvmet_subsys
*namespaces_to_subsys(
238 struct config_item
*item
)
240 return container_of(to_config_group(item
), struct nvmet_subsys
,
245 struct config_group group
;
248 static inline struct nvmet_host
*to_host(struct config_item
*item
)
250 return container_of(to_config_group(item
), struct nvmet_host
, group
);
253 static inline char *nvmet_host_name(struct nvmet_host
*host
)
255 return config_item_name(&host
->group
.cg_item
);
258 struct nvmet_host_link
{
259 struct list_head entry
;
260 struct nvmet_host
*host
;
263 struct nvmet_subsys_link
{
264 struct list_head entry
;
265 struct nvmet_subsys
*subsys
;
269 struct nvmet_fabrics_ops
{
270 struct module
*owner
;
273 bool has_keyed_sgls
: 1;
274 void (*queue_response
)(struct nvmet_req
*req
);
275 int (*add_port
)(struct nvmet_port
*port
);
276 void (*remove_port
)(struct nvmet_port
*port
);
277 void (*delete_ctrl
)(struct nvmet_ctrl
*ctrl
);
278 void (*disc_traddr
)(struct nvmet_req
*req
,
279 struct nvmet_port
*port
, char *traddr
);
280 u16 (*install_queue
)(struct nvmet_sq
*nvme_sq
);
281 void (*discovery_chg
)(struct nvmet_port
*port
);
284 #define NVMET_MAX_INLINE_BIOVEC 8
285 #define NVMET_MAX_INLINE_DATA_LEN NVMET_MAX_INLINE_BIOVEC * PAGE_SIZE
288 struct nvme_command
*cmd
;
289 struct nvme_completion
*cqe
;
293 struct scatterlist
*sg
;
294 struct bio_vec inline_bvec
[NVMET_MAX_INLINE_BIOVEC
];
297 struct bio inline_bio
;
302 struct bio_vec
*bvec
;
303 struct work_struct work
;
307 /* data length as parsed from the command: */
309 /* data length as parsed from the SGL descriptor: */
312 struct nvmet_port
*port
;
314 void (*execute
)(struct nvmet_req
*req
);
315 const struct nvmet_fabrics_ops
*ops
;
317 struct pci_dev
*p2p_dev
;
318 struct device
*p2p_client
;
323 extern struct workqueue_struct
*buffered_io_wq
;
325 static inline void nvmet_set_result(struct nvmet_req
*req
, u32 result
)
327 req
->cqe
->result
.u32
= cpu_to_le32(result
);
331 * NVMe command writes actually are DMA reads for us on the target side.
333 static inline enum dma_data_direction
334 nvmet_data_dir(struct nvmet_req
*req
)
336 return nvme_is_write(req
->cmd
) ? DMA_FROM_DEVICE
: DMA_TO_DEVICE
;
339 struct nvmet_async_event
{
340 struct list_head entry
;
346 static inline void nvmet_clear_aen_bit(struct nvmet_req
*req
, u32 bn
)
348 int rae
= le32_to_cpu(req
->cmd
->common
.cdw10
) & 1 << 15;
351 clear_bit(bn
, &req
->sq
->ctrl
->aen_masked
);
354 static inline bool nvmet_aen_bit_disabled(struct nvmet_ctrl
*ctrl
, u32 bn
)
356 if (!(READ_ONCE(ctrl
->aen_enabled
) & (1 << bn
)))
358 return test_and_set_bit(bn
, &ctrl
->aen_masked
);
361 void nvmet_get_feat_kato(struct nvmet_req
*req
);
362 void nvmet_get_feat_async_event(struct nvmet_req
*req
);
363 u16
nvmet_set_feat_kato(struct nvmet_req
*req
);
364 u16
nvmet_set_feat_async_event(struct nvmet_req
*req
, u32 mask
);
365 void nvmet_execute_async_event(struct nvmet_req
*req
);
367 u16
nvmet_parse_connect_cmd(struct nvmet_req
*req
);
368 void nvmet_bdev_set_limits(struct block_device
*bdev
, struct nvme_id_ns
*id
);
369 u16
nvmet_bdev_parse_io_cmd(struct nvmet_req
*req
);
370 u16
nvmet_file_parse_io_cmd(struct nvmet_req
*req
);
371 u16
nvmet_parse_admin_cmd(struct nvmet_req
*req
);
372 u16
nvmet_parse_discovery_cmd(struct nvmet_req
*req
);
373 u16
nvmet_parse_fabrics_cmd(struct nvmet_req
*req
);
375 bool nvmet_req_init(struct nvmet_req
*req
, struct nvmet_cq
*cq
,
376 struct nvmet_sq
*sq
, const struct nvmet_fabrics_ops
*ops
);
377 void nvmet_req_uninit(struct nvmet_req
*req
);
378 void nvmet_req_execute(struct nvmet_req
*req
);
379 void nvmet_req_complete(struct nvmet_req
*req
, u16 status
);
380 int nvmet_req_alloc_sgl(struct nvmet_req
*req
);
381 void nvmet_req_free_sgl(struct nvmet_req
*req
);
383 void nvmet_execute_keep_alive(struct nvmet_req
*req
);
385 void nvmet_cq_setup(struct nvmet_ctrl
*ctrl
, struct nvmet_cq
*cq
, u16 qid
,
387 void nvmet_sq_setup(struct nvmet_ctrl
*ctrl
, struct nvmet_sq
*sq
, u16 qid
,
389 void nvmet_sq_destroy(struct nvmet_sq
*sq
);
390 int nvmet_sq_init(struct nvmet_sq
*sq
);
392 void nvmet_ctrl_fatal_error(struct nvmet_ctrl
*ctrl
);
394 void nvmet_update_cc(struct nvmet_ctrl
*ctrl
, u32
new);
395 u16
nvmet_alloc_ctrl(const char *subsysnqn
, const char *hostnqn
,
396 struct nvmet_req
*req
, u32 kato
, struct nvmet_ctrl
**ctrlp
);
397 u16
nvmet_ctrl_find_get(const char *subsysnqn
, const char *hostnqn
, u16 cntlid
,
398 struct nvmet_req
*req
, struct nvmet_ctrl
**ret
);
399 void nvmet_ctrl_put(struct nvmet_ctrl
*ctrl
);
400 u16
nvmet_check_ctrl_status(struct nvmet_req
*req
, struct nvme_command
*cmd
);
402 struct nvmet_subsys
*nvmet_subsys_alloc(const char *subsysnqn
,
403 enum nvme_subsys_type type
);
404 void nvmet_subsys_put(struct nvmet_subsys
*subsys
);
405 void nvmet_subsys_del_ctrls(struct nvmet_subsys
*subsys
);
407 struct nvmet_ns
*nvmet_find_namespace(struct nvmet_ctrl
*ctrl
, __le32 nsid
);
408 void nvmet_put_namespace(struct nvmet_ns
*ns
);
409 int nvmet_ns_enable(struct nvmet_ns
*ns
);
410 void nvmet_ns_disable(struct nvmet_ns
*ns
);
411 struct nvmet_ns
*nvmet_ns_alloc(struct nvmet_subsys
*subsys
, u32 nsid
);
412 void nvmet_ns_free(struct nvmet_ns
*ns
);
414 void nvmet_send_ana_event(struct nvmet_subsys
*subsys
,
415 struct nvmet_port
*port
);
416 void nvmet_port_send_ana_event(struct nvmet_port
*port
);
418 int nvmet_register_transport(const struct nvmet_fabrics_ops
*ops
);
419 void nvmet_unregister_transport(const struct nvmet_fabrics_ops
*ops
);
421 int nvmet_enable_port(struct nvmet_port
*port
);
422 void nvmet_disable_port(struct nvmet_port
*port
);
424 void nvmet_referral_enable(struct nvmet_port
*parent
, struct nvmet_port
*port
);
425 void nvmet_referral_disable(struct nvmet_port
*parent
, struct nvmet_port
*port
);
427 u16
nvmet_copy_to_sgl(struct nvmet_req
*req
, off_t off
, const void *buf
,
429 u16
nvmet_copy_from_sgl(struct nvmet_req
*req
, off_t off
, void *buf
,
431 u16
nvmet_zero_sgl(struct nvmet_req
*req
, off_t off
, size_t len
);
433 u32
nvmet_get_log_page_len(struct nvme_command
*cmd
);
434 u64
nvmet_get_log_page_offset(struct nvme_command
*cmd
);
436 extern struct list_head
*nvmet_ports
;
437 void nvmet_port_disc_changed(struct nvmet_port
*port
,
438 struct nvmet_subsys
*subsys
);
439 void nvmet_subsys_disc_changed(struct nvmet_subsys
*subsys
,
440 struct nvmet_host
*host
);
441 void nvmet_add_async_event(struct nvmet_ctrl
*ctrl
, u8 event_type
,
442 u8 event_info
, u8 log_page
);
444 #define NVMET_QUEUE_SIZE 1024
445 #define NVMET_NR_QUEUES 128
446 #define NVMET_MAX_CMD NVMET_QUEUE_SIZE
449 * Nice round number that makes a list of nsids fit into a page.
450 * Should become tunable at some point in the future.
452 #define NVMET_MAX_NAMESPACES 1024
455 * 0 is not a valid ANA group ID, so we start numbering at 1.
457 * ANA Group 1 exists without manual intervention, has namespaces assigned to it
458 * by default, and is available in an optimized state through all ports.
460 #define NVMET_MAX_ANAGRPS 128
461 #define NVMET_DEFAULT_ANA_GRPID 1
464 #define NVMET_DISC_KATO_MS 120000
466 int __init
nvmet_init_configfs(void);
467 void __exit
nvmet_exit_configfs(void);
469 int __init
nvmet_init_discovery(void);
470 void nvmet_exit_discovery(void);
472 extern struct nvmet_subsys
*nvmet_disc_subsys
;
473 extern struct rw_semaphore nvmet_config_sem
;
475 extern u32 nvmet_ana_group_enabled
[NVMET_MAX_ANAGRPS
+ 1];
476 extern u64 nvmet_ana_chgcnt
;
477 extern struct rw_semaphore nvmet_ana_sem
;
479 bool nvmet_host_allowed(struct nvmet_subsys
*subsys
, const char *hostnqn
);
481 int nvmet_bdev_ns_enable(struct nvmet_ns
*ns
);
482 int nvmet_file_ns_enable(struct nvmet_ns
*ns
);
483 void nvmet_bdev_ns_disable(struct nvmet_ns
*ns
);
484 void nvmet_file_ns_disable(struct nvmet_ns
*ns
);
485 u16
nvmet_bdev_flush(struct nvmet_req
*req
);
486 u16
nvmet_file_flush(struct nvmet_req
*req
);
487 void nvmet_ns_changed(struct nvmet_subsys
*subsys
, u32 nsid
);
489 static inline u32
nvmet_rw_len(struct nvmet_req
*req
)
491 return ((u32
)le16_to_cpu(req
->cmd
->rw
.length
) + 1) <<
492 req
->ns
->blksize_shift
;
495 u16
errno_to_nvme_status(struct nvmet_req
*req
, int errno
);
497 /* Convert a 32-bit number to a 16-bit 0's based number */
498 static inline __le16
to0based(u32 a
)
500 return cpu_to_le16(max(1U, min(1U << 16, a
)) - 1);
503 #endif /* _NVMET_H */