xref: /openbmc/linux/drivers/nvme/host/nvme.h (revision e7bae9bb)
1 /* SPDX-License-Identifier: GPL-2.0 */
2 /*
3  * Copyright (c) 2011-2014, Intel Corporation.
4  */
5 
6 #ifndef _NVME_H
7 #define _NVME_H
8 
9 #include <linux/nvme.h>
10 #include <linux/cdev.h>
11 #include <linux/pci.h>
12 #include <linux/kref.h>
13 #include <linux/blk-mq.h>
14 #include <linux/lightnvm.h>
15 #include <linux/sed-opal.h>
16 #include <linux/fault-inject.h>
17 #include <linux/rcupdate.h>
18 #include <linux/wait.h>
19 #include <linux/t10-pi.h>
20 
21 #include <trace/events/block.h>
22 
23 extern unsigned int nvme_io_timeout;
24 #define NVME_IO_TIMEOUT	(nvme_io_timeout * HZ)
25 
26 extern unsigned int admin_timeout;
27 #define ADMIN_TIMEOUT	(admin_timeout * HZ)
28 
29 #define NVME_DEFAULT_KATO	5
30 #define NVME_KATO_GRACE		10
31 
32 #ifdef CONFIG_ARCH_NO_SG_CHAIN
33 #define  NVME_INLINE_SG_CNT  0
34 #define  NVME_INLINE_METADATA_SG_CNT  0
35 #else
36 #define  NVME_INLINE_SG_CNT  2
37 #define  NVME_INLINE_METADATA_SG_CNT  1
38 #endif
39 
40 /*
41  * Default to a 4K page size, with the intention to update this
42  * path in the future to accommodate architectures with differing
43  * kernel and IO page sizes.
44  */
45 #define NVME_CTRL_PAGE_SHIFT	12
46 #define NVME_CTRL_PAGE_SIZE	(1 << NVME_CTRL_PAGE_SHIFT)
47 
48 extern struct workqueue_struct *nvme_wq;
49 extern struct workqueue_struct *nvme_reset_wq;
50 extern struct workqueue_struct *nvme_delete_wq;
51 
52 enum {
53 	NVME_NS_LBA		= 0,
54 	NVME_NS_LIGHTNVM	= 1,
55 };
56 
57 /*
58  * List of workarounds for devices that required behavior not specified in
59  * the standard.
60  */
61 enum nvme_quirks {
62 	/*
63 	 * Prefers I/O aligned to a stripe size specified in a vendor
64 	 * specific Identify field.
65 	 */
66 	NVME_QUIRK_STRIPE_SIZE			= (1 << 0),
67 
68 	/*
69 	 * The controller doesn't handle Identify value others than 0 or 1
70 	 * correctly.
71 	 */
72 	NVME_QUIRK_IDENTIFY_CNS			= (1 << 1),
73 
74 	/*
75 	 * The controller deterministically returns O's on reads to
76 	 * logical blocks that deallocate was called on.
77 	 */
78 	NVME_QUIRK_DEALLOCATE_ZEROES		= (1 << 2),
79 
80 	/*
81 	 * The controller needs a delay before starts checking the device
82 	 * readiness, which is done by reading the NVME_CSTS_RDY bit.
83 	 */
84 	NVME_QUIRK_DELAY_BEFORE_CHK_RDY		= (1 << 3),
85 
86 	/*
87 	 * APST should not be used.
88 	 */
89 	NVME_QUIRK_NO_APST			= (1 << 4),
90 
91 	/*
92 	 * The deepest sleep state should not be used.
93 	 */
94 	NVME_QUIRK_NO_DEEPEST_PS		= (1 << 5),
95 
96 	/*
97 	 * Supports the LighNVM command set if indicated in vs[1].
98 	 */
99 	NVME_QUIRK_LIGHTNVM			= (1 << 6),
100 
101 	/*
102 	 * Set MEDIUM priority on SQ creation
103 	 */
104 	NVME_QUIRK_MEDIUM_PRIO_SQ		= (1 << 7),
105 
106 	/*
107 	 * Ignore device provided subnqn.
108 	 */
109 	NVME_QUIRK_IGNORE_DEV_SUBNQN		= (1 << 8),
110 
111 	/*
112 	 * Broken Write Zeroes.
113 	 */
114 	NVME_QUIRK_DISABLE_WRITE_ZEROES		= (1 << 9),
115 
116 	/*
117 	 * Force simple suspend/resume path.
118 	 */
119 	NVME_QUIRK_SIMPLE_SUSPEND		= (1 << 10),
120 
121 	/*
122 	 * Use only one interrupt vector for all queues
123 	 */
124 	NVME_QUIRK_SINGLE_VECTOR		= (1 << 11),
125 
126 	/*
127 	 * Use non-standard 128 bytes SQEs.
128 	 */
129 	NVME_QUIRK_128_BYTES_SQES		= (1 << 12),
130 
131 	/*
132 	 * Prevent tag overlap between queues
133 	 */
134 	NVME_QUIRK_SHARED_TAGS                  = (1 << 13),
135 
136 	/*
137 	 * Don't change the value of the temperature threshold feature
138 	 */
139 	NVME_QUIRK_NO_TEMP_THRESH_CHANGE	= (1 << 14),
140 
141 	/*
142 	 * The controller doesn't handle the Identify Namespace
143 	 * Identification Descriptor list subcommand despite claiming
144 	 * NVMe 1.3 compliance.
145 	 */
146 	NVME_QUIRK_NO_NS_DESC_LIST		= (1 << 15),
147 };
148 
149 /*
150  * Common request structure for NVMe passthrough.  All drivers must have
151  * this structure as the first member of their request-private data.
152  */
153 struct nvme_request {
154 	struct nvme_command	*cmd;
155 	union nvme_result	result;
156 	u8			retries;
157 	u8			flags;
158 	u16			status;
159 	struct nvme_ctrl	*ctrl;
160 };
161 
162 /*
163  * Mark a bio as coming in through the mpath node.
164  */
165 #define REQ_NVME_MPATH		REQ_DRV
166 
167 enum {
168 	NVME_REQ_CANCELLED		= (1 << 0),
169 	NVME_REQ_USERCMD		= (1 << 1),
170 };
171 
172 static inline struct nvme_request *nvme_req(struct request *req)
173 {
174 	return blk_mq_rq_to_pdu(req);
175 }
176 
177 static inline u16 nvme_req_qid(struct request *req)
178 {
179 	if (!req->rq_disk)
180 		return 0;
181 	return blk_mq_unique_tag_to_hwq(blk_mq_unique_tag(req)) + 1;
182 }
183 
184 /* The below value is the specific amount of delay needed before checking
185  * readiness in case of the PCI_DEVICE(0x1c58, 0x0003), which needs the
186  * NVME_QUIRK_DELAY_BEFORE_CHK_RDY quirk enabled. The value (in ms) was
187  * found empirically.
188  */
189 #define NVME_QUIRK_DELAY_AMOUNT		2300
190 
191 /*
192  * enum nvme_ctrl_state: Controller state
193  *
194  * @NVME_CTRL_NEW:		New controller just allocated, initial state
195  * @NVME_CTRL_LIVE:		Controller is connected and I/O capable
196  * @NVME_CTRL_RESETTING:	Controller is resetting (or scheduled reset)
197  * @NVME_CTRL_CONNECTING:	Controller is disconnected, now connecting the
198  *				transport
199  * @NVME_CTRL_DELETING:		Controller is deleting (or scheduled deletion)
200  * @NVME_CTRL_DELETING_NOIO:	Controller is deleting and I/O is not
201  *				disabled/failed immediately. This state comes
202  * 				after all async event processing took place and
203  * 				before ns removal and the controller deletion
204  * 				progress
205  * @NVME_CTRL_DEAD:		Controller is non-present/unresponsive during
206  *				shutdown or removal. In this case we forcibly
207  *				kill all inflight I/O as they have no chance to
208  *				complete
209  */
210 enum nvme_ctrl_state {
211 	NVME_CTRL_NEW,
212 	NVME_CTRL_LIVE,
213 	NVME_CTRL_RESETTING,
214 	NVME_CTRL_CONNECTING,
215 	NVME_CTRL_DELETING,
216 	NVME_CTRL_DELETING_NOIO,
217 	NVME_CTRL_DEAD,
218 };
219 
220 struct nvme_fault_inject {
221 #ifdef CONFIG_FAULT_INJECTION_DEBUG_FS
222 	struct fault_attr attr;
223 	struct dentry *parent;
224 	bool dont_retry;	/* DNR, do not retry */
225 	u16 status;		/* status code */
226 #endif
227 };
228 
229 struct nvme_cel {
230 	struct list_head	entry;
231 	struct nvme_effects_log	log;
232 	u8			csi;
233 };
234 
235 struct nvme_ctrl {
236 	bool comp_seen;
237 	enum nvme_ctrl_state state;
238 	bool identified;
239 	spinlock_t lock;
240 	struct mutex scan_lock;
241 	const struct nvme_ctrl_ops *ops;
242 	struct request_queue *admin_q;
243 	struct request_queue *connect_q;
244 	struct request_queue *fabrics_q;
245 	struct device *dev;
246 	int instance;
247 	int numa_node;
248 	struct blk_mq_tag_set *tagset;
249 	struct blk_mq_tag_set *admin_tagset;
250 	struct list_head namespaces;
251 	struct rw_semaphore namespaces_rwsem;
252 	struct device ctrl_device;
253 	struct device *device;	/* char device */
254 	struct cdev cdev;
255 	struct work_struct reset_work;
256 	struct work_struct delete_work;
257 	wait_queue_head_t state_wq;
258 
259 	struct nvme_subsystem *subsys;
260 	struct list_head subsys_entry;
261 
262 	struct opal_dev *opal_dev;
263 
264 	char name[12];
265 	u16 cntlid;
266 
267 	u32 ctrl_config;
268 	u16 mtfa;
269 	u32 queue_count;
270 
271 	u64 cap;
272 	u32 max_hw_sectors;
273 	u32 max_segments;
274 	u32 max_integrity_segments;
275 #ifdef CONFIG_BLK_DEV_ZONED
276 	u32 max_zone_append;
277 #endif
278 	u16 crdt[3];
279 	u16 oncs;
280 	u16 oacs;
281 	u16 nssa;
282 	u16 nr_streams;
283 	u16 sqsize;
284 	u32 max_namespaces;
285 	atomic_t abort_limit;
286 	u8 vwc;
287 	u32 vs;
288 	u32 sgls;
289 	u16 kas;
290 	u8 npss;
291 	u8 apsta;
292 	u16 wctemp;
293 	u16 cctemp;
294 	u32 oaes;
295 	u32 aen_result;
296 	u32 ctratt;
297 	unsigned int shutdown_timeout;
298 	unsigned int kato;
299 	bool subsystem;
300 	unsigned long quirks;
301 	struct nvme_id_power_state psd[32];
302 	struct nvme_effects_log *effects;
303 	struct list_head cels;
304 	struct work_struct scan_work;
305 	struct work_struct async_event_work;
306 	struct delayed_work ka_work;
307 	struct nvme_command ka_cmd;
308 	struct work_struct fw_act_work;
309 	unsigned long events;
310 
311 #ifdef CONFIG_NVME_MULTIPATH
312 	/* asymmetric namespace access: */
313 	u8 anacap;
314 	u8 anatt;
315 	u32 anagrpmax;
316 	u32 nanagrpid;
317 	struct mutex ana_lock;
318 	struct nvme_ana_rsp_hdr *ana_log_buf;
319 	size_t ana_log_size;
320 	struct timer_list anatt_timer;
321 	struct work_struct ana_work;
322 #endif
323 
324 	/* Power saving configuration */
325 	u64 ps_max_latency_us;
326 	bool apst_enabled;
327 
328 	/* PCIe only: */
329 	u32 hmpre;
330 	u32 hmmin;
331 	u32 hmminds;
332 	u16 hmmaxd;
333 
334 	/* Fabrics only */
335 	u32 ioccsz;
336 	u32 iorcsz;
337 	u16 icdoff;
338 	u16 maxcmd;
339 	int nr_reconnects;
340 	struct nvmf_ctrl_options *opts;
341 
342 	struct page *discard_page;
343 	unsigned long discard_page_busy;
344 
345 	struct nvme_fault_inject fault_inject;
346 };
347 
348 enum nvme_iopolicy {
349 	NVME_IOPOLICY_NUMA,
350 	NVME_IOPOLICY_RR,
351 };
352 
353 struct nvme_subsystem {
354 	int			instance;
355 	struct device		dev;
356 	/*
357 	 * Because we unregister the device on the last put we need
358 	 * a separate refcount.
359 	 */
360 	struct kref		ref;
361 	struct list_head	entry;
362 	struct mutex		lock;
363 	struct list_head	ctrls;
364 	struct list_head	nsheads;
365 	char			subnqn[NVMF_NQN_SIZE];
366 	char			serial[20];
367 	char			model[40];
368 	char			firmware_rev[8];
369 	u8			cmic;
370 	u16			vendor_id;
371 	u16			awupf;	/* 0's based awupf value. */
372 	struct ida		ns_ida;
373 #ifdef CONFIG_NVME_MULTIPATH
374 	enum nvme_iopolicy	iopolicy;
375 #endif
376 };
377 
378 /*
379  * Container structure for uniqueue namespace identifiers.
380  */
381 struct nvme_ns_ids {
382 	u8	eui64[8];
383 	u8	nguid[16];
384 	uuid_t	uuid;
385 	u8	csi;
386 };
387 
388 /*
389  * Anchor structure for namespaces.  There is one for each namespace in a
390  * NVMe subsystem that any of our controllers can see, and the namespace
391  * structure for each controller is chained of it.  For private namespaces
392  * there is a 1:1 relation to our namespace structures, that is ->list
393  * only ever has a single entry for private namespaces.
394  */
395 struct nvme_ns_head {
396 	struct list_head	list;
397 	struct srcu_struct      srcu;
398 	struct nvme_subsystem	*subsys;
399 	unsigned		ns_id;
400 	struct nvme_ns_ids	ids;
401 	struct list_head	entry;
402 	struct kref		ref;
403 	bool			shared;
404 	int			instance;
405 	struct nvme_effects_log *effects;
406 #ifdef CONFIG_NVME_MULTIPATH
407 	struct gendisk		*disk;
408 	struct bio_list		requeue_list;
409 	spinlock_t		requeue_lock;
410 	struct work_struct	requeue_work;
411 	struct mutex		lock;
412 	unsigned long		flags;
413 #define NVME_NSHEAD_DISK_LIVE	0
414 	struct nvme_ns __rcu	*current_path[];
415 #endif
416 };
417 
418 enum nvme_ns_features {
419 	NVME_NS_EXT_LBAS = 1 << 0, /* support extended LBA format */
420 	NVME_NS_METADATA_SUPPORTED = 1 << 1, /* support getting generated md */
421 };
422 
423 struct nvme_ns {
424 	struct list_head list;
425 
426 	struct nvme_ctrl *ctrl;
427 	struct request_queue *queue;
428 	struct gendisk *disk;
429 #ifdef CONFIG_NVME_MULTIPATH
430 	enum nvme_ana_state ana_state;
431 	u32 ana_grpid;
432 #endif
433 	struct list_head siblings;
434 	struct nvm_dev *ndev;
435 	struct kref kref;
436 	struct nvme_ns_head *head;
437 
438 	int lba_shift;
439 	u16 ms;
440 	u16 sgs;
441 	u32 sws;
442 	u8 pi_type;
443 #ifdef CONFIG_BLK_DEV_ZONED
444 	u64 zsze;
445 #endif
446 	unsigned long features;
447 	unsigned long flags;
448 #define NVME_NS_REMOVING	0
449 #define NVME_NS_DEAD     	1
450 #define NVME_NS_ANA_PENDING	2
451 
452 	struct nvme_fault_inject fault_inject;
453 
454 };
455 
456 /* NVMe ns supports metadata actions by the controller (generate/strip) */
457 static inline bool nvme_ns_has_pi(struct nvme_ns *ns)
458 {
459 	return ns->pi_type && ns->ms == sizeof(struct t10_pi_tuple);
460 }
461 
462 struct nvme_ctrl_ops {
463 	const char *name;
464 	struct module *module;
465 	unsigned int flags;
466 #define NVME_F_FABRICS			(1 << 0)
467 #define NVME_F_METADATA_SUPPORTED	(1 << 1)
468 #define NVME_F_PCI_P2PDMA		(1 << 2)
469 	int (*reg_read32)(struct nvme_ctrl *ctrl, u32 off, u32 *val);
470 	int (*reg_write32)(struct nvme_ctrl *ctrl, u32 off, u32 val);
471 	int (*reg_read64)(struct nvme_ctrl *ctrl, u32 off, u64 *val);
472 	void (*free_ctrl)(struct nvme_ctrl *ctrl);
473 	void (*submit_async_event)(struct nvme_ctrl *ctrl);
474 	void (*delete_ctrl)(struct nvme_ctrl *ctrl);
475 	int (*get_address)(struct nvme_ctrl *ctrl, char *buf, int size);
476 };
477 
478 #ifdef CONFIG_FAULT_INJECTION_DEBUG_FS
479 void nvme_fault_inject_init(struct nvme_fault_inject *fault_inj,
480 			    const char *dev_name);
481 void nvme_fault_inject_fini(struct nvme_fault_inject *fault_inject);
482 void nvme_should_fail(struct request *req);
483 #else
484 static inline void nvme_fault_inject_init(struct nvme_fault_inject *fault_inj,
485 					  const char *dev_name)
486 {
487 }
488 static inline void nvme_fault_inject_fini(struct nvme_fault_inject *fault_inj)
489 {
490 }
491 static inline void nvme_should_fail(struct request *req) {}
492 #endif
493 
494 static inline int nvme_reset_subsystem(struct nvme_ctrl *ctrl)
495 {
496 	if (!ctrl->subsystem)
497 		return -ENOTTY;
498 	return ctrl->ops->reg_write32(ctrl, NVME_REG_NSSR, 0x4E564D65);
499 }
500 
501 /*
502  * Convert a 512B sector number to a device logical block number.
503  */
504 static inline u64 nvme_sect_to_lba(struct nvme_ns *ns, sector_t sector)
505 {
506 	return sector >> (ns->lba_shift - SECTOR_SHIFT);
507 }
508 
509 /*
510  * Convert a device logical block number to a 512B sector number.
511  */
512 static inline sector_t nvme_lba_to_sect(struct nvme_ns *ns, u64 lba)
513 {
514 	return lba << (ns->lba_shift - SECTOR_SHIFT);
515 }
516 
517 /*
518  * Convert byte length to nvme's 0-based num dwords
519  */
520 static inline u32 nvme_bytes_to_numd(size_t len)
521 {
522 	return (len >> 2) - 1;
523 }
524 
525 static inline bool nvme_is_ana_error(u16 status)
526 {
527 	switch (status & 0x7ff) {
528 	case NVME_SC_ANA_TRANSITION:
529 	case NVME_SC_ANA_INACCESSIBLE:
530 	case NVME_SC_ANA_PERSISTENT_LOSS:
531 		return true;
532 	default:
533 		return false;
534 	}
535 }
536 
537 static inline bool nvme_is_path_error(u16 status)
538 {
539 	/* check for a status code type of 'path related status' */
540 	return (status & 0x700) == 0x300;
541 }
542 
543 /*
544  * Fill in the status and result information from the CQE, and then figure out
545  * if blk-mq will need to use IPI magic to complete the request, and if yes do
546  * so.  If not let the caller complete the request without an indirect function
547  * call.
548  */
549 static inline bool nvme_try_complete_req(struct request *req, __le16 status,
550 		union nvme_result result)
551 {
552 	struct nvme_request *rq = nvme_req(req);
553 
554 	rq->status = le16_to_cpu(status) >> 1;
555 	rq->result = result;
556 	/* inject error when permitted by fault injection framework */
557 	nvme_should_fail(req);
558 	if (unlikely(blk_should_fake_timeout(req->q)))
559 		return true;
560 	return blk_mq_complete_request_remote(req);
561 }
562 
563 static inline void nvme_get_ctrl(struct nvme_ctrl *ctrl)
564 {
565 	get_device(ctrl->device);
566 }
567 
568 static inline void nvme_put_ctrl(struct nvme_ctrl *ctrl)
569 {
570 	put_device(ctrl->device);
571 }
572 
573 static inline bool nvme_is_aen_req(u16 qid, __u16 command_id)
574 {
575 	return !qid && command_id >= NVME_AQ_BLK_MQ_DEPTH;
576 }
577 
578 void nvme_complete_rq(struct request *req);
579 bool nvme_cancel_request(struct request *req, void *data, bool reserved);
580 bool nvme_change_ctrl_state(struct nvme_ctrl *ctrl,
581 		enum nvme_ctrl_state new_state);
582 bool nvme_wait_reset(struct nvme_ctrl *ctrl);
583 int nvme_disable_ctrl(struct nvme_ctrl *ctrl);
584 int nvme_enable_ctrl(struct nvme_ctrl *ctrl);
585 int nvme_shutdown_ctrl(struct nvme_ctrl *ctrl);
586 int nvme_init_ctrl(struct nvme_ctrl *ctrl, struct device *dev,
587 		const struct nvme_ctrl_ops *ops, unsigned long quirks);
588 void nvme_uninit_ctrl(struct nvme_ctrl *ctrl);
589 void nvme_start_ctrl(struct nvme_ctrl *ctrl);
590 void nvme_stop_ctrl(struct nvme_ctrl *ctrl);
591 int nvme_init_identify(struct nvme_ctrl *ctrl);
592 
593 void nvme_remove_namespaces(struct nvme_ctrl *ctrl);
594 
595 int nvme_sec_submit(void *data, u16 spsp, u8 secp, void *buffer, size_t len,
596 		bool send);
597 
598 void nvme_complete_async_event(struct nvme_ctrl *ctrl, __le16 status,
599 		volatile union nvme_result *res);
600 
601 void nvme_stop_queues(struct nvme_ctrl *ctrl);
602 void nvme_start_queues(struct nvme_ctrl *ctrl);
603 void nvme_kill_queues(struct nvme_ctrl *ctrl);
604 void nvme_sync_queues(struct nvme_ctrl *ctrl);
605 void nvme_unfreeze(struct nvme_ctrl *ctrl);
606 void nvme_wait_freeze(struct nvme_ctrl *ctrl);
607 int nvme_wait_freeze_timeout(struct nvme_ctrl *ctrl, long timeout);
608 void nvme_start_freeze(struct nvme_ctrl *ctrl);
609 
610 #define NVME_QID_ANY -1
611 struct request *nvme_alloc_request(struct request_queue *q,
612 		struct nvme_command *cmd, blk_mq_req_flags_t flags, int qid);
613 void nvme_cleanup_cmd(struct request *req);
614 blk_status_t nvme_setup_cmd(struct nvme_ns *ns, struct request *req,
615 		struct nvme_command *cmd);
616 int nvme_submit_sync_cmd(struct request_queue *q, struct nvme_command *cmd,
617 		void *buf, unsigned bufflen);
618 int __nvme_submit_sync_cmd(struct request_queue *q, struct nvme_command *cmd,
619 		union nvme_result *result, void *buffer, unsigned bufflen,
620 		unsigned timeout, int qid, int at_head,
621 		blk_mq_req_flags_t flags, bool poll);
622 int nvme_set_features(struct nvme_ctrl *dev, unsigned int fid,
623 		      unsigned int dword11, void *buffer, size_t buflen,
624 		      u32 *result);
625 int nvme_get_features(struct nvme_ctrl *dev, unsigned int fid,
626 		      unsigned int dword11, void *buffer, size_t buflen,
627 		      u32 *result);
628 int nvme_set_queue_count(struct nvme_ctrl *ctrl, int *count);
629 void nvme_stop_keep_alive(struct nvme_ctrl *ctrl);
630 int nvme_reset_ctrl(struct nvme_ctrl *ctrl);
631 int nvme_reset_ctrl_sync(struct nvme_ctrl *ctrl);
632 int nvme_try_sched_reset(struct nvme_ctrl *ctrl);
633 int nvme_delete_ctrl(struct nvme_ctrl *ctrl);
634 
635 int nvme_get_log(struct nvme_ctrl *ctrl, u32 nsid, u8 log_page, u8 lsp, u8 csi,
636 		void *log, size_t size, u64 offset);
637 struct nvme_ns *nvme_get_ns_from_disk(struct gendisk *disk,
638 		struct nvme_ns_head **head, int *srcu_idx);
639 void nvme_put_ns_from_disk(struct nvme_ns_head *head, int idx);
640 
641 extern const struct attribute_group *nvme_ns_id_attr_groups[];
642 extern const struct block_device_operations nvme_ns_head_ops;
643 
644 #ifdef CONFIG_NVME_MULTIPATH
645 static inline bool nvme_ctrl_use_ana(struct nvme_ctrl *ctrl)
646 {
647 	return ctrl->ana_log_buf != NULL;
648 }
649 
650 void nvme_mpath_unfreeze(struct nvme_subsystem *subsys);
651 void nvme_mpath_wait_freeze(struct nvme_subsystem *subsys);
652 void nvme_mpath_start_freeze(struct nvme_subsystem *subsys);
653 void nvme_set_disk_name(char *disk_name, struct nvme_ns *ns,
654 			struct nvme_ctrl *ctrl, int *flags);
655 void nvme_failover_req(struct request *req);
656 void nvme_kick_requeue_lists(struct nvme_ctrl *ctrl);
657 int nvme_mpath_alloc_disk(struct nvme_ctrl *ctrl,struct nvme_ns_head *head);
658 void nvme_mpath_add_disk(struct nvme_ns *ns, struct nvme_id_ns *id);
659 void nvme_mpath_remove_disk(struct nvme_ns_head *head);
660 int nvme_mpath_init(struct nvme_ctrl *ctrl, struct nvme_id_ctrl *id);
661 void nvme_mpath_uninit(struct nvme_ctrl *ctrl);
662 void nvme_mpath_stop(struct nvme_ctrl *ctrl);
663 bool nvme_mpath_clear_current_path(struct nvme_ns *ns);
664 void nvme_mpath_clear_ctrl_paths(struct nvme_ctrl *ctrl);
665 struct nvme_ns *nvme_find_path(struct nvme_ns_head *head);
666 blk_qc_t nvme_ns_head_submit_bio(struct bio *bio);
667 
668 static inline void nvme_mpath_check_last_path(struct nvme_ns *ns)
669 {
670 	struct nvme_ns_head *head = ns->head;
671 
672 	if (head->disk && list_empty(&head->list))
673 		kblockd_schedule_work(&head->requeue_work);
674 }
675 
676 static inline void nvme_trace_bio_complete(struct request *req,
677         blk_status_t status)
678 {
679 	struct nvme_ns *ns = req->q->queuedata;
680 
681 	if (req->cmd_flags & REQ_NVME_MPATH)
682 		trace_block_bio_complete(ns->head->disk->queue, req->bio);
683 }
684 
685 static inline void nvme_mpath_update_disk_size(struct gendisk *disk)
686 {
687 	struct block_device *bdev = bdget_disk(disk, 0);
688 
689 	if (bdev) {
690 		bd_set_size(bdev, get_capacity(disk) << SECTOR_SHIFT);
691 		bdput(bdev);
692 	}
693 }
694 
695 extern struct device_attribute dev_attr_ana_grpid;
696 extern struct device_attribute dev_attr_ana_state;
697 extern struct device_attribute subsys_attr_iopolicy;
698 
699 #else
700 static inline bool nvme_ctrl_use_ana(struct nvme_ctrl *ctrl)
701 {
702 	return false;
703 }
704 /*
705  * Without the multipath code enabled, multiple controller per subsystems are
706  * visible as devices and thus we cannot use the subsystem instance.
707  */
708 static inline void nvme_set_disk_name(char *disk_name, struct nvme_ns *ns,
709 				      struct nvme_ctrl *ctrl, int *flags)
710 {
711 	sprintf(disk_name, "nvme%dn%d", ctrl->instance, ns->head->instance);
712 }
713 
714 static inline void nvme_failover_req(struct request *req)
715 {
716 }
717 static inline void nvme_kick_requeue_lists(struct nvme_ctrl *ctrl)
718 {
719 }
720 static inline int nvme_mpath_alloc_disk(struct nvme_ctrl *ctrl,
721 		struct nvme_ns_head *head)
722 {
723 	return 0;
724 }
725 static inline void nvme_mpath_add_disk(struct nvme_ns *ns,
726 		struct nvme_id_ns *id)
727 {
728 }
729 static inline void nvme_mpath_remove_disk(struct nvme_ns_head *head)
730 {
731 }
732 static inline bool nvme_mpath_clear_current_path(struct nvme_ns *ns)
733 {
734 	return false;
735 }
736 static inline void nvme_mpath_clear_ctrl_paths(struct nvme_ctrl *ctrl)
737 {
738 }
739 static inline void nvme_mpath_check_last_path(struct nvme_ns *ns)
740 {
741 }
742 static inline void nvme_trace_bio_complete(struct request *req,
743         blk_status_t status)
744 {
745 }
746 static inline int nvme_mpath_init(struct nvme_ctrl *ctrl,
747 		struct nvme_id_ctrl *id)
748 {
749 	if (ctrl->subsys->cmic & (1 << 3))
750 		dev_warn(ctrl->device,
751 "Please enable CONFIG_NVME_MULTIPATH for full support of multi-port devices.\n");
752 	return 0;
753 }
754 static inline void nvme_mpath_uninit(struct nvme_ctrl *ctrl)
755 {
756 }
757 static inline void nvme_mpath_stop(struct nvme_ctrl *ctrl)
758 {
759 }
760 static inline void nvme_mpath_unfreeze(struct nvme_subsystem *subsys)
761 {
762 }
763 static inline void nvme_mpath_wait_freeze(struct nvme_subsystem *subsys)
764 {
765 }
766 static inline void nvme_mpath_start_freeze(struct nvme_subsystem *subsys)
767 {
768 }
769 static inline void nvme_mpath_update_disk_size(struct gendisk *disk)
770 {
771 }
772 #endif /* CONFIG_NVME_MULTIPATH */
773 
774 #ifdef CONFIG_BLK_DEV_ZONED
775 int nvme_update_zone_info(struct gendisk *disk, struct nvme_ns *ns,
776 			  unsigned lbaf);
777 
778 int nvme_report_zones(struct gendisk *disk, sector_t sector,
779 		      unsigned int nr_zones, report_zones_cb cb, void *data);
780 
781 blk_status_t nvme_setup_zone_mgmt_send(struct nvme_ns *ns, struct request *req,
782 				       struct nvme_command *cmnd,
783 				       enum nvme_zone_mgmt_action action);
784 #else
785 #define nvme_report_zones NULL
786 
787 static inline blk_status_t nvme_setup_zone_mgmt_send(struct nvme_ns *ns,
788 		struct request *req, struct nvme_command *cmnd,
789 		enum nvme_zone_mgmt_action action)
790 {
791 	return BLK_STS_NOTSUPP;
792 }
793 
794 static inline int nvme_update_zone_info(struct gendisk *disk,
795 					struct nvme_ns *ns,
796 					unsigned lbaf)
797 {
798 	dev_warn(ns->ctrl->device,
799 		 "Please enable CONFIG_BLK_DEV_ZONED to support ZNS devices\n");
800 	return -EPROTONOSUPPORT;
801 }
802 #endif
803 
804 #ifdef CONFIG_NVM
805 int nvme_nvm_register(struct nvme_ns *ns, char *disk_name, int node);
806 void nvme_nvm_unregister(struct nvme_ns *ns);
807 extern const struct attribute_group nvme_nvm_attr_group;
808 int nvme_nvm_ioctl(struct nvme_ns *ns, unsigned int cmd, unsigned long arg);
809 #else
810 static inline int nvme_nvm_register(struct nvme_ns *ns, char *disk_name,
811 				    int node)
812 {
813 	return 0;
814 }
815 
816 static inline void nvme_nvm_unregister(struct nvme_ns *ns) {};
817 static inline int nvme_nvm_ioctl(struct nvme_ns *ns, unsigned int cmd,
818 							unsigned long arg)
819 {
820 	return -ENOTTY;
821 }
822 #endif /* CONFIG_NVM */
823 
824 static inline struct nvme_ns *nvme_get_ns_from_dev(struct device *dev)
825 {
826 	return dev_to_disk(dev)->private_data;
827 }
828 
829 #ifdef CONFIG_NVME_HWMON
830 int nvme_hwmon_init(struct nvme_ctrl *ctrl);
831 #else
832 static inline int nvme_hwmon_init(struct nvme_ctrl *ctrl)
833 {
834 	return 0;
835 }
836 #endif
837 
838 u32 nvme_command_effects(struct nvme_ctrl *ctrl, struct nvme_ns *ns,
839 			 u8 opcode);
840 void nvme_execute_passthru_rq(struct request *rq);
841 struct nvme_ctrl *nvme_ctrl_get_by_path(const char *path);
842 struct nvme_ns *nvme_find_get_ns(struct nvme_ctrl *ctrl, unsigned nsid);
843 void nvme_put_ns(struct nvme_ns *ns);
844 
845 #endif /* _NVME_H */
846