xref: /openbmc/linux/drivers/nvme/host/nvme.h (revision 276e552e)
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 NVME_ADMIN_TIMEOUT	(admin_timeout * HZ)
28 
29 #define NVME_DEFAULT_KATO	5
30 
31 #ifdef CONFIG_ARCH_NO_SG_CHAIN
32 #define  NVME_INLINE_SG_CNT  0
33 #define  NVME_INLINE_METADATA_SG_CNT  0
34 #else
35 #define  NVME_INLINE_SG_CNT  2
36 #define  NVME_INLINE_METADATA_SG_CNT  1
37 #endif
38 
39 /*
40  * Default to a 4K page size, with the intention to update this
41  * path in the future to accommodate architectures with differing
42  * kernel and IO page sizes.
43  */
44 #define NVME_CTRL_PAGE_SHIFT	12
45 #define NVME_CTRL_PAGE_SIZE	(1 << NVME_CTRL_PAGE_SHIFT)
46 
47 extern struct workqueue_struct *nvme_wq;
48 extern struct workqueue_struct *nvme_reset_wq;
49 extern struct workqueue_struct *nvme_delete_wq;
50 
51 enum {
52 	NVME_NS_LBA		= 0,
53 	NVME_NS_LIGHTNVM	= 1,
54 };
55 
56 /*
57  * List of workarounds for devices that required behavior not specified in
58  * the standard.
59  */
60 enum nvme_quirks {
61 	/*
62 	 * Prefers I/O aligned to a stripe size specified in a vendor
63 	 * specific Identify field.
64 	 */
65 	NVME_QUIRK_STRIPE_SIZE			= (1 << 0),
66 
67 	/*
68 	 * The controller doesn't handle Identify value others than 0 or 1
69 	 * correctly.
70 	 */
71 	NVME_QUIRK_IDENTIFY_CNS			= (1 << 1),
72 
73 	/*
74 	 * The controller deterministically returns O's on reads to
75 	 * logical blocks that deallocate was called on.
76 	 */
77 	NVME_QUIRK_DEALLOCATE_ZEROES		= (1 << 2),
78 
79 	/*
80 	 * The controller needs a delay before starts checking the device
81 	 * readiness, which is done by reading the NVME_CSTS_RDY bit.
82 	 */
83 	NVME_QUIRK_DELAY_BEFORE_CHK_RDY		= (1 << 3),
84 
85 	/*
86 	 * APST should not be used.
87 	 */
88 	NVME_QUIRK_NO_APST			= (1 << 4),
89 
90 	/*
91 	 * The deepest sleep state should not be used.
92 	 */
93 	NVME_QUIRK_NO_DEEPEST_PS		= (1 << 5),
94 
95 	/*
96 	 * Supports the LighNVM command set if indicated in vs[1].
97 	 */
98 	NVME_QUIRK_LIGHTNVM			= (1 << 6),
99 
100 	/*
101 	 * Set MEDIUM priority on SQ creation
102 	 */
103 	NVME_QUIRK_MEDIUM_PRIO_SQ		= (1 << 7),
104 
105 	/*
106 	 * Ignore device provided subnqn.
107 	 */
108 	NVME_QUIRK_IGNORE_DEV_SUBNQN		= (1 << 8),
109 
110 	/*
111 	 * Broken Write Zeroes.
112 	 */
113 	NVME_QUIRK_DISABLE_WRITE_ZEROES		= (1 << 9),
114 
115 	/*
116 	 * Force simple suspend/resume path.
117 	 */
118 	NVME_QUIRK_SIMPLE_SUSPEND		= (1 << 10),
119 
120 	/*
121 	 * Use only one interrupt vector for all queues
122 	 */
123 	NVME_QUIRK_SINGLE_VECTOR		= (1 << 11),
124 
125 	/*
126 	 * Use non-standard 128 bytes SQEs.
127 	 */
128 	NVME_QUIRK_128_BYTES_SQES		= (1 << 12),
129 
130 	/*
131 	 * Prevent tag overlap between queues
132 	 */
133 	NVME_QUIRK_SHARED_TAGS                  = (1 << 13),
134 
135 	/*
136 	 * Don't change the value of the temperature threshold feature
137 	 */
138 	NVME_QUIRK_NO_TEMP_THRESH_CHANGE	= (1 << 14),
139 
140 	/*
141 	 * The controller doesn't handle the Identify Namespace
142 	 * Identification Descriptor list subcommand despite claiming
143 	 * NVMe 1.3 compliance.
144 	 */
145 	NVME_QUIRK_NO_NS_DESC_LIST		= (1 << 15),
146 
147 	/*
148 	 * The controller does not properly handle DMA addresses over
149 	 * 48 bits.
150 	 */
151 	NVME_QUIRK_DMA_ADDRESS_BITS_48		= (1 << 16),
152 };
153 
154 /*
155  * Common request structure for NVMe passthrough.  All drivers must have
156  * this structure as the first member of their request-private data.
157  */
158 struct nvme_request {
159 	struct nvme_command	*cmd;
160 	union nvme_result	result;
161 	u8			retries;
162 	u8			flags;
163 	u16			status;
164 	struct nvme_ctrl	*ctrl;
165 };
166 
167 /*
168  * Mark a bio as coming in through the mpath node.
169  */
170 #define REQ_NVME_MPATH		REQ_DRV
171 
172 enum {
173 	NVME_REQ_CANCELLED		= (1 << 0),
174 	NVME_REQ_USERCMD		= (1 << 1),
175 };
176 
177 static inline struct nvme_request *nvme_req(struct request *req)
178 {
179 	return blk_mq_rq_to_pdu(req);
180 }
181 
182 static inline u16 nvme_req_qid(struct request *req)
183 {
184 	if (!req->q->queuedata)
185 		return 0;
186 
187 	return req->mq_hctx->queue_num + 1;
188 }
189 
190 /* The below value is the specific amount of delay needed before checking
191  * readiness in case of the PCI_DEVICE(0x1c58, 0x0003), which needs the
192  * NVME_QUIRK_DELAY_BEFORE_CHK_RDY quirk enabled. The value (in ms) was
193  * found empirically.
194  */
195 #define NVME_QUIRK_DELAY_AMOUNT		2300
196 
197 /*
198  * enum nvme_ctrl_state: Controller state
199  *
200  * @NVME_CTRL_NEW:		New controller just allocated, initial state
201  * @NVME_CTRL_LIVE:		Controller is connected and I/O capable
202  * @NVME_CTRL_RESETTING:	Controller is resetting (or scheduled reset)
203  * @NVME_CTRL_CONNECTING:	Controller is disconnected, now connecting the
204  *				transport
205  * @NVME_CTRL_DELETING:		Controller is deleting (or scheduled deletion)
206  * @NVME_CTRL_DELETING_NOIO:	Controller is deleting and I/O is not
207  *				disabled/failed immediately. This state comes
208  * 				after all async event processing took place and
209  * 				before ns removal and the controller deletion
210  * 				progress
211  * @NVME_CTRL_DEAD:		Controller is non-present/unresponsive during
212  *				shutdown or removal. In this case we forcibly
213  *				kill all inflight I/O as they have no chance to
214  *				complete
215  */
216 enum nvme_ctrl_state {
217 	NVME_CTRL_NEW,
218 	NVME_CTRL_LIVE,
219 	NVME_CTRL_RESETTING,
220 	NVME_CTRL_CONNECTING,
221 	NVME_CTRL_DELETING,
222 	NVME_CTRL_DELETING_NOIO,
223 	NVME_CTRL_DEAD,
224 };
225 
226 struct nvme_fault_inject {
227 #ifdef CONFIG_FAULT_INJECTION_DEBUG_FS
228 	struct fault_attr attr;
229 	struct dentry *parent;
230 	bool dont_retry;	/* DNR, do not retry */
231 	u16 status;		/* status code */
232 #endif
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 #ifdef CONFIG_NVME_HWMON
255 	struct device *hwmon_device;
256 #endif
257 	struct cdev cdev;
258 	struct work_struct reset_work;
259 	struct work_struct delete_work;
260 	wait_queue_head_t state_wq;
261 
262 	struct nvme_subsystem *subsys;
263 	struct list_head subsys_entry;
264 
265 	struct opal_dev *opal_dev;
266 
267 	char name[12];
268 	u16 cntlid;
269 
270 	u32 ctrl_config;
271 	u16 mtfa;
272 	u32 queue_count;
273 
274 	u64 cap;
275 	u32 max_hw_sectors;
276 	u32 max_segments;
277 	u32 max_integrity_segments;
278 	u32 max_discard_sectors;
279 	u32 max_discard_segments;
280 	u32 max_zeroes_sectors;
281 #ifdef CONFIG_BLK_DEV_ZONED
282 	u32 max_zone_append;
283 #endif
284 	u16 crdt[3];
285 	u16 oncs;
286 	u16 oacs;
287 	u16 nssa;
288 	u16 nr_streams;
289 	u16 sqsize;
290 	u32 max_namespaces;
291 	atomic_t abort_limit;
292 	u8 vwc;
293 	u32 vs;
294 	u32 sgls;
295 	u16 kas;
296 	u8 npss;
297 	u8 apsta;
298 	u16 wctemp;
299 	u16 cctemp;
300 	u32 oaes;
301 	u32 aen_result;
302 	u32 ctratt;
303 	unsigned int shutdown_timeout;
304 	unsigned int kato;
305 	bool subsystem;
306 	unsigned long quirks;
307 	struct nvme_id_power_state psd[32];
308 	struct nvme_effects_log *effects;
309 	struct xarray cels;
310 	struct work_struct scan_work;
311 	struct work_struct async_event_work;
312 	struct delayed_work ka_work;
313 	struct delayed_work failfast_work;
314 	struct nvme_command ka_cmd;
315 	struct work_struct fw_act_work;
316 	unsigned long events;
317 
318 #ifdef CONFIG_NVME_MULTIPATH
319 	/* asymmetric namespace access: */
320 	u8 anacap;
321 	u8 anatt;
322 	u32 anagrpmax;
323 	u32 nanagrpid;
324 	struct mutex ana_lock;
325 	struct nvme_ana_rsp_hdr *ana_log_buf;
326 	size_t ana_log_size;
327 	struct timer_list anatt_timer;
328 	struct work_struct ana_work;
329 #endif
330 
331 	/* Power saving configuration */
332 	u64 ps_max_latency_us;
333 	bool apst_enabled;
334 
335 	/* PCIe only: */
336 	u32 hmpre;
337 	u32 hmmin;
338 	u32 hmminds;
339 	u16 hmmaxd;
340 
341 	/* Fabrics only */
342 	u32 ioccsz;
343 	u32 iorcsz;
344 	u16 icdoff;
345 	u16 maxcmd;
346 	int nr_reconnects;
347 	unsigned long flags;
348 #define NVME_CTRL_FAILFAST_EXPIRED	0
349 	struct nvmf_ctrl_options *opts;
350 
351 	struct page *discard_page;
352 	unsigned long discard_page_busy;
353 
354 	struct nvme_fault_inject fault_inject;
355 };
356 
357 enum nvme_iopolicy {
358 	NVME_IOPOLICY_NUMA,
359 	NVME_IOPOLICY_RR,
360 };
361 
362 struct nvme_subsystem {
363 	int			instance;
364 	struct device		dev;
365 	/*
366 	 * Because we unregister the device on the last put we need
367 	 * a separate refcount.
368 	 */
369 	struct kref		ref;
370 	struct list_head	entry;
371 	struct mutex		lock;
372 	struct list_head	ctrls;
373 	struct list_head	nsheads;
374 	char			subnqn[NVMF_NQN_SIZE];
375 	char			serial[20];
376 	char			model[40];
377 	char			firmware_rev[8];
378 	u8			cmic;
379 	u16			vendor_id;
380 	u16			awupf;	/* 0's based awupf value. */
381 	struct ida		ns_ida;
382 #ifdef CONFIG_NVME_MULTIPATH
383 	enum nvme_iopolicy	iopolicy;
384 #endif
385 };
386 
387 /*
388  * Container structure for uniqueue namespace identifiers.
389  */
390 struct nvme_ns_ids {
391 	u8	eui64[8];
392 	u8	nguid[16];
393 	uuid_t	uuid;
394 	u8	csi;
395 };
396 
397 /*
398  * Anchor structure for namespaces.  There is one for each namespace in a
399  * NVMe subsystem that any of our controllers can see, and the namespace
400  * structure for each controller is chained of it.  For private namespaces
401  * there is a 1:1 relation to our namespace structures, that is ->list
402  * only ever has a single entry for private namespaces.
403  */
404 struct nvme_ns_head {
405 	struct list_head	list;
406 	struct srcu_struct      srcu;
407 	struct nvme_subsystem	*subsys;
408 	unsigned		ns_id;
409 	struct nvme_ns_ids	ids;
410 	struct list_head	entry;
411 	struct kref		ref;
412 	bool			shared;
413 	int			instance;
414 	struct nvme_effects_log *effects;
415 
416 	struct cdev		cdev;
417 	struct device		cdev_device;
418 
419 	struct gendisk		*disk;
420 #ifdef CONFIG_NVME_MULTIPATH
421 	struct bio_list		requeue_list;
422 	spinlock_t		requeue_lock;
423 	struct work_struct	requeue_work;
424 	struct mutex		lock;
425 	unsigned long		flags;
426 #define NVME_NSHEAD_DISK_LIVE	0
427 	struct nvme_ns __rcu	*current_path[];
428 #endif
429 };
430 
431 static inline bool nvme_ns_head_multipath(struct nvme_ns_head *head)
432 {
433 	return IS_ENABLED(CONFIG_NVME_MULTIPATH) && head->disk;
434 }
435 
436 enum nvme_ns_features {
437 	NVME_NS_EXT_LBAS = 1 << 0, /* support extended LBA format */
438 	NVME_NS_METADATA_SUPPORTED = 1 << 1, /* support getting generated md */
439 };
440 
441 struct nvme_ns {
442 	struct list_head list;
443 
444 	struct nvme_ctrl *ctrl;
445 	struct request_queue *queue;
446 	struct gendisk *disk;
447 #ifdef CONFIG_NVME_MULTIPATH
448 	enum nvme_ana_state ana_state;
449 	u32 ana_grpid;
450 #endif
451 	struct list_head siblings;
452 	struct nvm_dev *ndev;
453 	struct kref kref;
454 	struct nvme_ns_head *head;
455 
456 	int lba_shift;
457 	u16 ms;
458 	u16 sgs;
459 	u32 sws;
460 	u8 pi_type;
461 #ifdef CONFIG_BLK_DEV_ZONED
462 	u64 zsze;
463 #endif
464 	unsigned long features;
465 	unsigned long flags;
466 #define NVME_NS_REMOVING	0
467 #define NVME_NS_DEAD     	1
468 #define NVME_NS_ANA_PENDING	2
469 #define NVME_NS_FORCE_RO	3
470 
471 	struct cdev		cdev;
472 	struct device		cdev_device;
473 
474 	struct nvme_fault_inject fault_inject;
475 
476 };
477 
478 /* NVMe ns supports metadata actions by the controller (generate/strip) */
479 static inline bool nvme_ns_has_pi(struct nvme_ns *ns)
480 {
481 	return ns->pi_type && ns->ms == sizeof(struct t10_pi_tuple);
482 }
483 
484 struct nvme_ctrl_ops {
485 	const char *name;
486 	struct module *module;
487 	unsigned int flags;
488 #define NVME_F_FABRICS			(1 << 0)
489 #define NVME_F_METADATA_SUPPORTED	(1 << 1)
490 #define NVME_F_PCI_P2PDMA		(1 << 2)
491 	int (*reg_read32)(struct nvme_ctrl *ctrl, u32 off, u32 *val);
492 	int (*reg_write32)(struct nvme_ctrl *ctrl, u32 off, u32 val);
493 	int (*reg_read64)(struct nvme_ctrl *ctrl, u32 off, u64 *val);
494 	void (*free_ctrl)(struct nvme_ctrl *ctrl);
495 	void (*submit_async_event)(struct nvme_ctrl *ctrl);
496 	void (*delete_ctrl)(struct nvme_ctrl *ctrl);
497 	int (*get_address)(struct nvme_ctrl *ctrl, char *buf, int size);
498 };
499 
500 #ifdef CONFIG_FAULT_INJECTION_DEBUG_FS
501 void nvme_fault_inject_init(struct nvme_fault_inject *fault_inj,
502 			    const char *dev_name);
503 void nvme_fault_inject_fini(struct nvme_fault_inject *fault_inject);
504 void nvme_should_fail(struct request *req);
505 #else
506 static inline void nvme_fault_inject_init(struct nvme_fault_inject *fault_inj,
507 					  const char *dev_name)
508 {
509 }
510 static inline void nvme_fault_inject_fini(struct nvme_fault_inject *fault_inj)
511 {
512 }
513 static inline void nvme_should_fail(struct request *req) {}
514 #endif
515 
516 static inline int nvme_reset_subsystem(struct nvme_ctrl *ctrl)
517 {
518 	if (!ctrl->subsystem)
519 		return -ENOTTY;
520 	return ctrl->ops->reg_write32(ctrl, NVME_REG_NSSR, 0x4E564D65);
521 }
522 
523 /*
524  * Convert a 512B sector number to a device logical block number.
525  */
526 static inline u64 nvme_sect_to_lba(struct nvme_ns *ns, sector_t sector)
527 {
528 	return sector >> (ns->lba_shift - SECTOR_SHIFT);
529 }
530 
531 /*
532  * Convert a device logical block number to a 512B sector number.
533  */
534 static inline sector_t nvme_lba_to_sect(struct nvme_ns *ns, u64 lba)
535 {
536 	return lba << (ns->lba_shift - SECTOR_SHIFT);
537 }
538 
539 /*
540  * Convert byte length to nvme's 0-based num dwords
541  */
542 static inline u32 nvme_bytes_to_numd(size_t len)
543 {
544 	return (len >> 2) - 1;
545 }
546 
547 static inline bool nvme_is_ana_error(u16 status)
548 {
549 	switch (status & 0x7ff) {
550 	case NVME_SC_ANA_TRANSITION:
551 	case NVME_SC_ANA_INACCESSIBLE:
552 	case NVME_SC_ANA_PERSISTENT_LOSS:
553 		return true;
554 	default:
555 		return false;
556 	}
557 }
558 
559 static inline bool nvme_is_path_error(u16 status)
560 {
561 	/* check for a status code type of 'path related status' */
562 	return (status & 0x700) == 0x300;
563 }
564 
565 /*
566  * Fill in the status and result information from the CQE, and then figure out
567  * if blk-mq will need to use IPI magic to complete the request, and if yes do
568  * so.  If not let the caller complete the request without an indirect function
569  * call.
570  */
571 static inline bool nvme_try_complete_req(struct request *req, __le16 status,
572 		union nvme_result result)
573 {
574 	struct nvme_request *rq = nvme_req(req);
575 
576 	rq->status = le16_to_cpu(status) >> 1;
577 	rq->result = result;
578 	/* inject error when permitted by fault injection framework */
579 	nvme_should_fail(req);
580 	if (unlikely(blk_should_fake_timeout(req->q)))
581 		return true;
582 	return blk_mq_complete_request_remote(req);
583 }
584 
585 static inline void nvme_get_ctrl(struct nvme_ctrl *ctrl)
586 {
587 	get_device(ctrl->device);
588 }
589 
590 static inline void nvme_put_ctrl(struct nvme_ctrl *ctrl)
591 {
592 	put_device(ctrl->device);
593 }
594 
595 static inline bool nvme_is_aen_req(u16 qid, __u16 command_id)
596 {
597 	return !qid && command_id >= NVME_AQ_BLK_MQ_DEPTH;
598 }
599 
600 void nvme_complete_rq(struct request *req);
601 blk_status_t nvme_host_path_error(struct request *req);
602 bool nvme_cancel_request(struct request *req, void *data, bool reserved);
603 void nvme_cancel_tagset(struct nvme_ctrl *ctrl);
604 void nvme_cancel_admin_tagset(struct nvme_ctrl *ctrl);
605 bool nvme_change_ctrl_state(struct nvme_ctrl *ctrl,
606 		enum nvme_ctrl_state new_state);
607 bool nvme_wait_reset(struct nvme_ctrl *ctrl);
608 int nvme_disable_ctrl(struct nvme_ctrl *ctrl);
609 int nvme_enable_ctrl(struct nvme_ctrl *ctrl);
610 int nvme_shutdown_ctrl(struct nvme_ctrl *ctrl);
611 int nvme_init_ctrl(struct nvme_ctrl *ctrl, struct device *dev,
612 		const struct nvme_ctrl_ops *ops, unsigned long quirks);
613 void nvme_uninit_ctrl(struct nvme_ctrl *ctrl);
614 void nvme_start_ctrl(struct nvme_ctrl *ctrl);
615 void nvme_stop_ctrl(struct nvme_ctrl *ctrl);
616 int nvme_init_ctrl_finish(struct nvme_ctrl *ctrl);
617 
618 void nvme_remove_namespaces(struct nvme_ctrl *ctrl);
619 
620 int nvme_sec_submit(void *data, u16 spsp, u8 secp, void *buffer, size_t len,
621 		bool send);
622 
623 void nvme_complete_async_event(struct nvme_ctrl *ctrl, __le16 status,
624 		volatile union nvme_result *res);
625 
626 void nvme_stop_queues(struct nvme_ctrl *ctrl);
627 void nvme_start_queues(struct nvme_ctrl *ctrl);
628 void nvme_kill_queues(struct nvme_ctrl *ctrl);
629 void nvme_sync_queues(struct nvme_ctrl *ctrl);
630 void nvme_sync_io_queues(struct nvme_ctrl *ctrl);
631 void nvme_unfreeze(struct nvme_ctrl *ctrl);
632 void nvme_wait_freeze(struct nvme_ctrl *ctrl);
633 int nvme_wait_freeze_timeout(struct nvme_ctrl *ctrl, long timeout);
634 void nvme_start_freeze(struct nvme_ctrl *ctrl);
635 
636 #define NVME_QID_ANY -1
637 struct request *nvme_alloc_request(struct request_queue *q,
638 		struct nvme_command *cmd, blk_mq_req_flags_t flags);
639 void nvme_cleanup_cmd(struct request *req);
640 blk_status_t nvme_setup_cmd(struct nvme_ns *ns, struct request *req);
641 blk_status_t nvme_fail_nonready_command(struct nvme_ctrl *ctrl,
642 		struct request *req);
643 bool __nvme_check_ready(struct nvme_ctrl *ctrl, struct request *rq,
644 		bool queue_live);
645 
646 static inline bool nvme_check_ready(struct nvme_ctrl *ctrl, struct request *rq,
647 		bool queue_live)
648 {
649 	if (likely(ctrl->state == NVME_CTRL_LIVE))
650 		return true;
651 	if (ctrl->ops->flags & NVME_F_FABRICS &&
652 	    ctrl->state == NVME_CTRL_DELETING)
653 		return true;
654 	return __nvme_check_ready(ctrl, rq, queue_live);
655 }
656 int nvme_submit_sync_cmd(struct request_queue *q, struct nvme_command *cmd,
657 		void *buf, unsigned bufflen);
658 int __nvme_submit_sync_cmd(struct request_queue *q, struct nvme_command *cmd,
659 		union nvme_result *result, void *buffer, unsigned bufflen,
660 		unsigned timeout, int qid, int at_head,
661 		blk_mq_req_flags_t flags, bool poll);
662 int nvme_set_features(struct nvme_ctrl *dev, unsigned int fid,
663 		      unsigned int dword11, void *buffer, size_t buflen,
664 		      u32 *result);
665 int nvme_get_features(struct nvme_ctrl *dev, unsigned int fid,
666 		      unsigned int dword11, void *buffer, size_t buflen,
667 		      u32 *result);
668 int nvme_set_queue_count(struct nvme_ctrl *ctrl, int *count);
669 void nvme_stop_keep_alive(struct nvme_ctrl *ctrl);
670 int nvme_reset_ctrl(struct nvme_ctrl *ctrl);
671 int nvme_reset_ctrl_sync(struct nvme_ctrl *ctrl);
672 int nvme_try_sched_reset(struct nvme_ctrl *ctrl);
673 int nvme_delete_ctrl(struct nvme_ctrl *ctrl);
674 void nvme_queue_scan(struct nvme_ctrl *ctrl);
675 int nvme_get_log(struct nvme_ctrl *ctrl, u32 nsid, u8 log_page, u8 lsp, u8 csi,
676 		void *log, size_t size, u64 offset);
677 struct nvme_ns *nvme_get_ns_from_disk(struct gendisk *disk,
678 		struct nvme_ns_head **head, int *srcu_idx);
679 void nvme_put_ns_from_disk(struct nvme_ns_head *head, int idx);
680 bool nvme_tryget_ns_head(struct nvme_ns_head *head);
681 void nvme_put_ns_head(struct nvme_ns_head *head);
682 int nvme_cdev_add(struct cdev *cdev, struct device *cdev_device,
683 		const struct file_operations *fops, struct module *owner);
684 void nvme_cdev_del(struct cdev *cdev, struct device *cdev_device);
685 int nvme_ioctl(struct block_device *bdev, fmode_t mode,
686 		unsigned int cmd, unsigned long arg);
687 long nvme_ns_chr_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
688 int nvme_ns_head_ioctl(struct block_device *bdev, fmode_t mode,
689 		unsigned int cmd, unsigned long arg);
690 long nvme_ns_head_chr_ioctl(struct file *file, unsigned int cmd,
691 		unsigned long arg);
692 long nvme_dev_ioctl(struct file *file, unsigned int cmd,
693 		unsigned long arg);
694 int nvme_getgeo(struct block_device *bdev, struct hd_geometry *geo);
695 
696 extern const struct attribute_group *nvme_ns_id_attr_groups[];
697 extern const struct pr_ops nvme_pr_ops;
698 extern const struct block_device_operations nvme_ns_head_ops;
699 
700 #ifdef CONFIG_NVME_MULTIPATH
701 static inline bool nvme_ctrl_use_ana(struct nvme_ctrl *ctrl)
702 {
703 	return ctrl->ana_log_buf != NULL;
704 }
705 
706 void nvme_mpath_unfreeze(struct nvme_subsystem *subsys);
707 void nvme_mpath_wait_freeze(struct nvme_subsystem *subsys);
708 void nvme_mpath_start_freeze(struct nvme_subsystem *subsys);
709 bool nvme_mpath_set_disk_name(struct nvme_ns *ns, char *disk_name, int *flags);
710 void nvme_failover_req(struct request *req);
711 void nvme_kick_requeue_lists(struct nvme_ctrl *ctrl);
712 int nvme_mpath_alloc_disk(struct nvme_ctrl *ctrl,struct nvme_ns_head *head);
713 void nvme_mpath_add_disk(struct nvme_ns *ns, struct nvme_id_ns *id);
714 void nvme_mpath_remove_disk(struct nvme_ns_head *head);
715 int nvme_mpath_init(struct nvme_ctrl *ctrl, struct nvme_id_ctrl *id);
716 void nvme_mpath_uninit(struct nvme_ctrl *ctrl);
717 void nvme_mpath_stop(struct nvme_ctrl *ctrl);
718 bool nvme_mpath_clear_current_path(struct nvme_ns *ns);
719 void nvme_mpath_clear_ctrl_paths(struct nvme_ctrl *ctrl);
720 struct nvme_ns *nvme_find_path(struct nvme_ns_head *head);
721 
722 static inline void nvme_mpath_check_last_path(struct nvme_ns *ns)
723 {
724 	struct nvme_ns_head *head = ns->head;
725 
726 	if (head->disk && list_empty(&head->list))
727 		kblockd_schedule_work(&head->requeue_work);
728 }
729 
730 static inline void nvme_trace_bio_complete(struct request *req)
731 {
732 	struct nvme_ns *ns = req->q->queuedata;
733 
734 	if (req->cmd_flags & REQ_NVME_MPATH)
735 		trace_block_bio_complete(ns->head->disk->queue, req->bio);
736 }
737 
738 extern struct device_attribute dev_attr_ana_grpid;
739 extern struct device_attribute dev_attr_ana_state;
740 extern struct device_attribute subsys_attr_iopolicy;
741 
742 #else
743 static inline bool nvme_ctrl_use_ana(struct nvme_ctrl *ctrl)
744 {
745 	return false;
746 }
747 static inline bool nvme_mpath_set_disk_name(struct nvme_ns *ns, char *disk_name,
748 		int *flags)
749 {
750 	return false;
751 }
752 static inline void nvme_failover_req(struct request *req)
753 {
754 }
755 static inline void nvme_kick_requeue_lists(struct nvme_ctrl *ctrl)
756 {
757 }
758 static inline int nvme_mpath_alloc_disk(struct nvme_ctrl *ctrl,
759 		struct nvme_ns_head *head)
760 {
761 	return 0;
762 }
763 static inline void nvme_mpath_add_disk(struct nvme_ns *ns,
764 		struct nvme_id_ns *id)
765 {
766 }
767 static inline void nvme_mpath_remove_disk(struct nvme_ns_head *head)
768 {
769 }
770 static inline bool nvme_mpath_clear_current_path(struct nvme_ns *ns)
771 {
772 	return false;
773 }
774 static inline void nvme_mpath_clear_ctrl_paths(struct nvme_ctrl *ctrl)
775 {
776 }
777 static inline void nvme_mpath_check_last_path(struct nvme_ns *ns)
778 {
779 }
780 static inline void nvme_trace_bio_complete(struct request *req)
781 {
782 }
783 static inline int nvme_mpath_init(struct nvme_ctrl *ctrl,
784 		struct nvme_id_ctrl *id)
785 {
786 	if (ctrl->subsys->cmic & NVME_CTRL_CMIC_ANA)
787 		dev_warn(ctrl->device,
788 "Please enable CONFIG_NVME_MULTIPATH for full support of multi-port devices.\n");
789 	return 0;
790 }
791 static inline void nvme_mpath_uninit(struct nvme_ctrl *ctrl)
792 {
793 }
794 static inline void nvme_mpath_stop(struct nvme_ctrl *ctrl)
795 {
796 }
797 static inline void nvme_mpath_unfreeze(struct nvme_subsystem *subsys)
798 {
799 }
800 static inline void nvme_mpath_wait_freeze(struct nvme_subsystem *subsys)
801 {
802 }
803 static inline void nvme_mpath_start_freeze(struct nvme_subsystem *subsys)
804 {
805 }
806 #endif /* CONFIG_NVME_MULTIPATH */
807 
808 int nvme_revalidate_zones(struct nvme_ns *ns);
809 #ifdef CONFIG_BLK_DEV_ZONED
810 int nvme_update_zone_info(struct nvme_ns *ns, unsigned lbaf);
811 int nvme_report_zones(struct gendisk *disk, sector_t sector,
812 		      unsigned int nr_zones, report_zones_cb cb, void *data);
813 
814 blk_status_t nvme_setup_zone_mgmt_send(struct nvme_ns *ns, struct request *req,
815 				       struct nvme_command *cmnd,
816 				       enum nvme_zone_mgmt_action action);
817 #else
818 #define nvme_report_zones NULL
819 
820 static inline blk_status_t nvme_setup_zone_mgmt_send(struct nvme_ns *ns,
821 		struct request *req, struct nvme_command *cmnd,
822 		enum nvme_zone_mgmt_action action)
823 {
824 	return BLK_STS_NOTSUPP;
825 }
826 
827 static inline int nvme_update_zone_info(struct nvme_ns *ns, unsigned lbaf)
828 {
829 	dev_warn(ns->ctrl->device,
830 		 "Please enable CONFIG_BLK_DEV_ZONED to support ZNS devices\n");
831 	return -EPROTONOSUPPORT;
832 }
833 #endif
834 
835 #ifdef CONFIG_NVM
836 int nvme_nvm_register(struct nvme_ns *ns, char *disk_name, int node);
837 void nvme_nvm_unregister(struct nvme_ns *ns);
838 extern const struct attribute_group nvme_nvm_attr_group;
839 int nvme_nvm_ioctl(struct nvme_ns *ns, unsigned int cmd, void __user *argp);
840 #else
841 static inline int nvme_nvm_register(struct nvme_ns *ns, char *disk_name,
842 				    int node)
843 {
844 	return 0;
845 }
846 
847 static inline void nvme_nvm_unregister(struct nvme_ns *ns) {};
848 static inline int nvme_nvm_ioctl(struct nvme_ns *ns, unsigned int cmd,
849 		void __user *argp)
850 {
851 	return -ENOTTY;
852 }
853 #endif /* CONFIG_NVM */
854 
855 static inline struct nvme_ns *nvme_get_ns_from_dev(struct device *dev)
856 {
857 	return dev_to_disk(dev)->private_data;
858 }
859 
860 #ifdef CONFIG_NVME_HWMON
861 int nvme_hwmon_init(struct nvme_ctrl *ctrl);
862 void nvme_hwmon_exit(struct nvme_ctrl *ctrl);
863 #else
864 static inline int nvme_hwmon_init(struct nvme_ctrl *ctrl)
865 {
866 	return 0;
867 }
868 
869 static inline void nvme_hwmon_exit(struct nvme_ctrl *ctrl)
870 {
871 }
872 #endif
873 
874 u32 nvme_command_effects(struct nvme_ctrl *ctrl, struct nvme_ns *ns,
875 			 u8 opcode);
876 void nvme_execute_passthru_rq(struct request *rq);
877 struct nvme_ctrl *nvme_ctrl_from_file(struct file *file);
878 struct nvme_ns *nvme_find_get_ns(struct nvme_ctrl *ctrl, unsigned nsid);
879 void nvme_put_ns(struct nvme_ns *ns);
880 
881 #endif /* _NVME_H */
882