2 * Copyright (c) 2015-2016 HGST, a Western Digital Company.
4 * This program is free software; you can redistribute it and/or modify it
5 * under the terms and conditions of the GNU General Public License,
6 * version 2, as published by the Free Software Foundation.
8 * This program is distributed in the hope it will be useful, but WITHOUT
9 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
10 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
17 #include <linux/dma-mapping.h>
18 #include <linux/types.h>
19 #include <linux/device.h>
20 #include <linux/kref.h>
21 #include <linux/percpu-refcount.h>
22 #include <linux/list.h>
23 #include <linux/mutex.h>
24 #include <linux/uuid.h>
25 #include <linux/nvme.h>
26 #include <linux/configfs.h>
27 #include <linux/rcupdate.h>
28 #include <linux/blkdev.h>
29 #include <linux/radix-tree.h>
31 #define NVMET_ASYNC_EVENTS 4
32 #define NVMET_ERROR_LOG_SLOTS 128
33 #define NVMET_NO_ERROR_LOC ((u16)-1)
36 * Supported optional AENs:
38 #define NVMET_AEN_CFG_OPTIONAL \
39 (NVME_AEN_CFG_NS_ATTR | NVME_AEN_CFG_ANA_CHANGE)
40 #define NVMET_DISC_AEN_CFG_OPTIONAL \
41 (NVME_AEN_CFG_DISC_CHANGE)
44 * Plus mandatory SMART AENs (we'll never send them, but allow enabling them):
46 #define NVMET_AEN_CFG_ALL \
47 (NVME_SMART_CRIT_SPARE | NVME_SMART_CRIT_TEMPERATURE | \
48 NVME_SMART_CRIT_RELIABILITY | NVME_SMART_CRIT_MEDIA | \
49 NVME_SMART_CRIT_VOLATILE_MEMORY | NVMET_AEN_CFG_OPTIONAL)
51 /* Helper Macros when NVMe error is NVME_SC_CONNECT_INVALID_PARAM
52 * The 16 bit shift is to set IATTR bit to 1, which means offending
53 * offset starts in the data section of connect()
55 #define IPO_IATTR_CONNECT_DATA(x) \
56 (cpu_to_le32((1 << 16) | (offsetof(struct nvmf_connect_data, x))))
57 #define IPO_IATTR_CONNECT_SQE(x) \
58 (cpu_to_le32(offsetof(struct nvmf_connect_command, x)))
61 struct list_head dev_link;
62 struct percpu_ref ref;
63 struct block_device *bdev;
75 struct nvmet_subsys *subsys;
76 const char *device_path;
78 struct config_group device_group;
79 struct config_group group;
81 struct completion disable_done;
83 struct kmem_cache *bvec_cache;
86 struct pci_dev *p2p_dev;
89 static inline struct nvmet_ns *to_nvmet_ns(struct config_item *item)
91 return container_of(to_config_group(item), struct nvmet_ns, group);
94 static inline struct device *nvmet_ns_dev(struct nvmet_ns *ns)
96 return ns->bdev ? disk_to_dev(ns->bdev->bd_disk) : NULL;
105 struct nvmet_ctrl *ctrl;
106 struct percpu_ref ref;
111 struct completion free_done;
112 struct completion confirm_done;
115 struct nvmet_ana_group {
116 struct config_group group;
117 struct nvmet_port *port;
121 static inline struct nvmet_ana_group *to_ana_group(struct config_item *item)
123 return container_of(to_config_group(item), struct nvmet_ana_group,
128 * struct nvmet_port - Common structure to keep port
129 * information for the target.
130 * @entry: Entry into referrals or transport list.
131 * @disc_addr: Address information is stored in a format defined
132 * for a discovery log page entry.
133 * @group: ConfigFS group for this element's folder.
134 * @priv: Private data for the transport.
137 struct list_head entry;
138 struct nvmf_disc_rsp_page_entry disc_addr;
139 struct config_group group;
140 struct config_group subsys_group;
141 struct list_head subsystems;
142 struct config_group referrals_group;
143 struct list_head referrals;
144 struct list_head global_entry;
145 struct config_group ana_groups_group;
146 struct nvmet_ana_group ana_default_group;
147 enum nvme_ana_state *ana_state;
150 int inline_data_size;
153 static inline struct nvmet_port *to_nvmet_port(struct config_item *item)
155 return container_of(to_config_group(item), struct nvmet_port,
159 static inline struct nvmet_port *ana_groups_to_port(
160 struct config_item *item)
162 return container_of(to_config_group(item), struct nvmet_port,
167 struct nvmet_subsys *subsys;
168 struct nvmet_cq **cqs;
169 struct nvmet_sq **sqs;
182 struct nvmet_port *port;
185 unsigned long aen_masked;
186 struct nvmet_req *async_event_cmds[NVMET_ASYNC_EVENTS];
187 unsigned int nr_async_event_cmds;
188 struct list_head async_events;
189 struct work_struct async_event_work;
191 struct list_head subsys_entry;
193 struct delayed_work ka_work;
194 struct work_struct fatal_err_work;
196 const struct nvmet_fabrics_ops *ops;
198 __le32 *changed_ns_list;
201 char subsysnqn[NVMF_NQN_FIELD_LEN];
202 char hostnqn[NVMF_NQN_FIELD_LEN];
204 struct device *p2p_client;
205 struct radix_tree_root p2p_ns_map;
207 spinlock_t error_lock;
209 struct nvme_error_slot slots[NVMET_ERROR_LOG_SLOTS];
212 struct nvmet_subsys {
213 enum nvme_subsys_type type;
218 struct list_head namespaces;
219 unsigned int nr_namespaces;
220 unsigned int max_nsid;
222 struct list_head ctrls;
224 struct list_head hosts;
233 struct config_group group;
235 struct config_group namespaces_group;
236 struct config_group allowed_hosts_group;
239 static inline struct nvmet_subsys *to_subsys(struct config_item *item)
241 return container_of(to_config_group(item), struct nvmet_subsys, group);
244 static inline struct nvmet_subsys *namespaces_to_subsys(
245 struct config_item *item)
247 return container_of(to_config_group(item), struct nvmet_subsys,
252 struct config_group group;
255 static inline struct nvmet_host *to_host(struct config_item *item)
257 return container_of(to_config_group(item), struct nvmet_host, group);
260 static inline char *nvmet_host_name(struct nvmet_host *host)
262 return config_item_name(&host->group.cg_item);
265 struct nvmet_host_link {
266 struct list_head entry;
267 struct nvmet_host *host;
270 struct nvmet_subsys_link {
271 struct list_head entry;
272 struct nvmet_subsys *subsys;
276 struct nvmet_fabrics_ops {
277 struct module *owner;
280 bool has_keyed_sgls : 1;
281 void (*queue_response)(struct nvmet_req *req);
282 int (*add_port)(struct nvmet_port *port);
283 void (*remove_port)(struct nvmet_port *port);
284 void (*delete_ctrl)(struct nvmet_ctrl *ctrl);
285 void (*disc_traddr)(struct nvmet_req *req,
286 struct nvmet_port *port, char *traddr);
287 u16 (*install_queue)(struct nvmet_sq *nvme_sq);
290 #define NVMET_MAX_INLINE_BIOVEC 8
291 #define NVMET_MAX_INLINE_DATA_LEN NVMET_MAX_INLINE_BIOVEC * PAGE_SIZE
294 struct nvme_command *cmd;
295 struct nvme_completion *rsp;
299 struct scatterlist *sg;
300 struct bio_vec inline_bvec[NVMET_MAX_INLINE_BIOVEC];
303 struct bio inline_bio;
308 struct bio_vec *bvec;
309 struct work_struct work;
313 /* data length as parsed from the command: */
315 /* data length as parsed from the SGL descriptor: */
318 struct nvmet_port *port;
320 void (*execute)(struct nvmet_req *req);
321 const struct nvmet_fabrics_ops *ops;
323 struct pci_dev *p2p_dev;
324 struct device *p2p_client;
329 extern struct workqueue_struct *buffered_io_wq;
331 static inline void nvmet_set_result(struct nvmet_req *req, u32 result)
333 req->rsp->result.u32 = cpu_to_le32(result);
337 * NVMe command writes actually are DMA reads for us on the target side.
339 static inline enum dma_data_direction
340 nvmet_data_dir(struct nvmet_req *req)
342 return nvme_is_write(req->cmd) ? DMA_FROM_DEVICE : DMA_TO_DEVICE;
345 struct nvmet_async_event {
346 struct list_head entry;
352 static inline void nvmet_clear_aen_bit(struct nvmet_req *req, u32 bn)
354 int rae = le32_to_cpu(req->cmd->common.cdw10) & 1 << 15;
357 clear_bit(bn, &req->sq->ctrl->aen_masked);
360 static inline bool nvmet_aen_bit_disabled(struct nvmet_ctrl *ctrl, u32 bn)
362 if (!(READ_ONCE(ctrl->aen_enabled) & (1 << bn)))
364 return test_and_set_bit(bn, &ctrl->aen_masked);
367 void nvmet_get_feat_kato(struct nvmet_req *req);
368 void nvmet_get_feat_async_event(struct nvmet_req *req);
369 u16 nvmet_set_feat_kato(struct nvmet_req *req);
370 u16 nvmet_set_feat_async_event(struct nvmet_req *req, u32 mask);
371 void nvmet_execute_async_event(struct nvmet_req *req);
373 u16 nvmet_parse_connect_cmd(struct nvmet_req *req);
374 u16 nvmet_bdev_parse_io_cmd(struct nvmet_req *req);
375 u16 nvmet_file_parse_io_cmd(struct nvmet_req *req);
376 u16 nvmet_parse_admin_cmd(struct nvmet_req *req);
377 u16 nvmet_parse_discovery_cmd(struct nvmet_req *req);
378 u16 nvmet_parse_fabrics_cmd(struct nvmet_req *req);
380 bool nvmet_req_init(struct nvmet_req *req, struct nvmet_cq *cq,
381 struct nvmet_sq *sq, const struct nvmet_fabrics_ops *ops);
382 void nvmet_req_uninit(struct nvmet_req *req);
383 void nvmet_req_execute(struct nvmet_req *req);
384 void nvmet_req_complete(struct nvmet_req *req, u16 status);
385 int nvmet_req_alloc_sgl(struct nvmet_req *req);
386 void nvmet_req_free_sgl(struct nvmet_req *req);
388 void nvmet_execute_keep_alive(struct nvmet_req *req);
390 void nvmet_cq_setup(struct nvmet_ctrl *ctrl, struct nvmet_cq *cq, u16 qid,
392 void nvmet_sq_setup(struct nvmet_ctrl *ctrl, struct nvmet_sq *sq, u16 qid,
394 void nvmet_sq_destroy(struct nvmet_sq *sq);
395 int nvmet_sq_init(struct nvmet_sq *sq);
397 void nvmet_ctrl_fatal_error(struct nvmet_ctrl *ctrl);
399 void nvmet_update_cc(struct nvmet_ctrl *ctrl, u32 new);
400 u16 nvmet_alloc_ctrl(const char *subsysnqn, const char *hostnqn,
401 struct nvmet_req *req, u32 kato, struct nvmet_ctrl **ctrlp);
402 u16 nvmet_ctrl_find_get(const char *subsysnqn, const char *hostnqn, u16 cntlid,
403 struct nvmet_req *req, struct nvmet_ctrl **ret);
404 void nvmet_ctrl_put(struct nvmet_ctrl *ctrl);
405 u16 nvmet_check_ctrl_status(struct nvmet_req *req, struct nvme_command *cmd);
407 struct nvmet_subsys *nvmet_subsys_alloc(const char *subsysnqn,
408 enum nvme_subsys_type type);
409 void nvmet_subsys_put(struct nvmet_subsys *subsys);
410 void nvmet_subsys_del_ctrls(struct nvmet_subsys *subsys);
412 struct nvmet_ns *nvmet_find_namespace(struct nvmet_ctrl *ctrl, __le32 nsid);
413 void nvmet_put_namespace(struct nvmet_ns *ns);
414 int nvmet_ns_enable(struct nvmet_ns *ns);
415 void nvmet_ns_disable(struct nvmet_ns *ns);
416 struct nvmet_ns *nvmet_ns_alloc(struct nvmet_subsys *subsys, u32 nsid);
417 void nvmet_ns_free(struct nvmet_ns *ns);
419 void nvmet_send_ana_event(struct nvmet_subsys *subsys,
420 struct nvmet_port *port);
421 void nvmet_port_send_ana_event(struct nvmet_port *port);
423 int nvmet_register_transport(const struct nvmet_fabrics_ops *ops);
424 void nvmet_unregister_transport(const struct nvmet_fabrics_ops *ops);
426 int nvmet_enable_port(struct nvmet_port *port);
427 void nvmet_disable_port(struct nvmet_port *port);
429 void nvmet_referral_enable(struct nvmet_port *parent, struct nvmet_port *port);
430 void nvmet_referral_disable(struct nvmet_port *parent, struct nvmet_port *port);
432 u16 nvmet_copy_to_sgl(struct nvmet_req *req, off_t off, const void *buf,
434 u16 nvmet_copy_from_sgl(struct nvmet_req *req, off_t off, void *buf,
436 u16 nvmet_zero_sgl(struct nvmet_req *req, off_t off, size_t len);
438 u32 nvmet_get_log_page_len(struct nvme_command *cmd);
440 extern struct list_head *nvmet_ports;
441 void nvmet_port_disc_changed(struct nvmet_port *port,
442 struct nvmet_subsys *subsys);
443 void nvmet_subsys_disc_changed(struct nvmet_subsys *subsys,
444 struct nvmet_host *host);
445 void nvmet_add_async_event(struct nvmet_ctrl *ctrl, u8 event_type,
446 u8 event_info, u8 log_page);
448 #define NVMET_QUEUE_SIZE 1024
449 #define NVMET_NR_QUEUES 128
450 #define NVMET_MAX_CMD NVMET_QUEUE_SIZE
453 * Nice round number that makes a list of nsids fit into a page.
454 * Should become tunable at some point in the future.
456 #define NVMET_MAX_NAMESPACES 1024
459 * 0 is not a valid ANA group ID, so we start numbering at 1.
461 * ANA Group 1 exists without manual intervention, has namespaces assigned to it
462 * by default, and is available in an optimized state through all ports.
464 #define NVMET_MAX_ANAGRPS 128
465 #define NVMET_DEFAULT_ANA_GRPID 1
468 #define NVMET_DISC_KATO_MS 120000
470 int __init nvmet_init_configfs(void);
471 void __exit nvmet_exit_configfs(void);
473 int __init nvmet_init_discovery(void);
474 void nvmet_exit_discovery(void);
476 extern struct nvmet_subsys *nvmet_disc_subsys;
477 extern struct rw_semaphore nvmet_config_sem;
479 extern u32 nvmet_ana_group_enabled[NVMET_MAX_ANAGRPS + 1];
480 extern u64 nvmet_ana_chgcnt;
481 extern struct rw_semaphore nvmet_ana_sem;
483 bool nvmet_host_allowed(struct nvmet_subsys *subsys, const char *hostnqn);
485 int nvmet_bdev_ns_enable(struct nvmet_ns *ns);
486 int nvmet_file_ns_enable(struct nvmet_ns *ns);
487 void nvmet_bdev_ns_disable(struct nvmet_ns *ns);
488 void nvmet_file_ns_disable(struct nvmet_ns *ns);
489 u16 nvmet_bdev_flush(struct nvmet_req *req);
490 u16 nvmet_file_flush(struct nvmet_req *req);
491 void nvmet_ns_changed(struct nvmet_subsys *subsys, u32 nsid);
493 static inline u32 nvmet_rw_len(struct nvmet_req *req)
495 return ((u32)le16_to_cpu(req->cmd->rw.length) + 1) <<
496 req->ns->blksize_shift;
499 u16 errno_to_nvme_status(struct nvmet_req *req, int errno);
500 #endif /* _NVMET_H */