1 /* SPDX-License-Identifier: GPL-2.0 */
3 * Copyright (c) 2011-2014, Intel Corporation.
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/sed-opal.h>
15 #include <linux/fault-inject.h>
16 #include <linux/rcupdate.h>
17 #include <linux/wait.h>
18 #include <linux/t10-pi.h>
20 #include <trace/events/block.h>
22 extern unsigned int nvme_io_timeout;
23 #define NVME_IO_TIMEOUT (nvme_io_timeout * HZ)
25 extern unsigned int admin_timeout;
26 #define NVME_ADMIN_TIMEOUT (admin_timeout * HZ)
28 #define NVME_DEFAULT_KATO 5
30 #ifdef CONFIG_ARCH_NO_SG_CHAIN
31 #define NVME_INLINE_SG_CNT 0
32 #define NVME_INLINE_METADATA_SG_CNT 0
34 #define NVME_INLINE_SG_CNT 2
35 #define NVME_INLINE_METADATA_SG_CNT 1
39 * Default to a 4K page size, with the intention to update this
40 * path in the future to accommodate architectures with differing
41 * kernel and IO page sizes.
43 #define NVME_CTRL_PAGE_SHIFT 12
44 #define NVME_CTRL_PAGE_SIZE (1 << NVME_CTRL_PAGE_SHIFT)
46 extern struct workqueue_struct *nvme_wq;
47 extern struct workqueue_struct *nvme_reset_wq;
48 extern struct workqueue_struct *nvme_delete_wq;
51 * List of workarounds for devices that required behavior not specified in
56 * Prefers I/O aligned to a stripe size specified in a vendor
57 * specific Identify field.
59 NVME_QUIRK_STRIPE_SIZE = (1 << 0),
62 * The controller doesn't handle Identify value others than 0 or 1
65 NVME_QUIRK_IDENTIFY_CNS = (1 << 1),
68 * The controller deterministically returns O's on reads to
69 * logical blocks that deallocate was called on.
71 NVME_QUIRK_DEALLOCATE_ZEROES = (1 << 2),
74 * The controller needs a delay before starts checking the device
75 * readiness, which is done by reading the NVME_CSTS_RDY bit.
77 NVME_QUIRK_DELAY_BEFORE_CHK_RDY = (1 << 3),
80 * APST should not be used.
82 NVME_QUIRK_NO_APST = (1 << 4),
85 * The deepest sleep state should not be used.
87 NVME_QUIRK_NO_DEEPEST_PS = (1 << 5),
90 * Set MEDIUM priority on SQ creation
92 NVME_QUIRK_MEDIUM_PRIO_SQ = (1 << 7),
95 * Ignore device provided subnqn.
97 NVME_QUIRK_IGNORE_DEV_SUBNQN = (1 << 8),
100 * Broken Write Zeroes.
102 NVME_QUIRK_DISABLE_WRITE_ZEROES = (1 << 9),
105 * Force simple suspend/resume path.
107 NVME_QUIRK_SIMPLE_SUSPEND = (1 << 10),
110 * Use only one interrupt vector for all queues
112 NVME_QUIRK_SINGLE_VECTOR = (1 << 11),
115 * Use non-standard 128 bytes SQEs.
117 NVME_QUIRK_128_BYTES_SQES = (1 << 12),
120 * Prevent tag overlap between queues
122 NVME_QUIRK_SHARED_TAGS = (1 << 13),
125 * Don't change the value of the temperature threshold feature
127 NVME_QUIRK_NO_TEMP_THRESH_CHANGE = (1 << 14),
130 * The controller doesn't handle the Identify Namespace
131 * Identification Descriptor list subcommand despite claiming
132 * NVMe 1.3 compliance.
134 NVME_QUIRK_NO_NS_DESC_LIST = (1 << 15),
137 * The controller does not properly handle DMA addresses over
140 NVME_QUIRK_DMA_ADDRESS_BITS_48 = (1 << 16),
143 * The controller requires the command_id value be limited, so skip
144 * encoding the generation sequence number.
146 NVME_QUIRK_SKIP_CID_GEN = (1 << 17),
149 * Reports garbage in the namespace identifiers (eui64, nguid, uuid).
151 NVME_QUIRK_BOGUS_NID = (1 << 18),
154 * No temperature thresholds for channels other than 0 (Composite).
156 NVME_QUIRK_NO_SECONDARY_TEMP_THRESH = (1 << 19),
160 * Common request structure for NVMe passthrough. All drivers must have
161 * this structure as the first member of their request-private data.
163 struct nvme_request {
164 struct nvme_command *cmd;
165 union nvme_result result;
170 #ifdef CONFIG_NVME_MULTIPATH
171 unsigned long start_time;
173 struct nvme_ctrl *ctrl;
177 * Mark a bio as coming in through the mpath node.
179 #define REQ_NVME_MPATH REQ_DRV
182 NVME_REQ_CANCELLED = (1 << 0),
183 NVME_REQ_USERCMD = (1 << 1),
184 NVME_MPATH_IO_STATS = (1 << 2),
187 static inline struct nvme_request *nvme_req(struct request *req)
189 return blk_mq_rq_to_pdu(req);
192 static inline u16 nvme_req_qid(struct request *req)
194 if (!req->q->queuedata)
197 return req->mq_hctx->queue_num + 1;
200 /* The below value is the specific amount of delay needed before checking
201 * readiness in case of the PCI_DEVICE(0x1c58, 0x0003), which needs the
202 * NVME_QUIRK_DELAY_BEFORE_CHK_RDY quirk enabled. The value (in ms) was
205 #define NVME_QUIRK_DELAY_AMOUNT 2300
208 * enum nvme_ctrl_state: Controller state
210 * @NVME_CTRL_NEW: New controller just allocated, initial state
211 * @NVME_CTRL_LIVE: Controller is connected and I/O capable
212 * @NVME_CTRL_RESETTING: Controller is resetting (or scheduled reset)
213 * @NVME_CTRL_CONNECTING: Controller is disconnected, now connecting the
215 * @NVME_CTRL_DELETING: Controller is deleting (or scheduled deletion)
216 * @NVME_CTRL_DELETING_NOIO: Controller is deleting and I/O is not
217 * disabled/failed immediately. This state comes
218 * after all async event processing took place and
219 * before ns removal and the controller deletion
221 * @NVME_CTRL_DEAD: Controller is non-present/unresponsive during
222 * shutdown or removal. In this case we forcibly
223 * kill all inflight I/O as they have no chance to
226 enum nvme_ctrl_state {
230 NVME_CTRL_CONNECTING,
232 NVME_CTRL_DELETING_NOIO,
236 struct nvme_fault_inject {
237 #ifdef CONFIG_FAULT_INJECTION_DEBUG_FS
238 struct fault_attr attr;
239 struct dentry *parent;
240 bool dont_retry; /* DNR, do not retry */
241 u16 status; /* status code */
245 enum nvme_ctrl_flags {
246 NVME_CTRL_FAILFAST_EXPIRED = 0,
247 NVME_CTRL_ADMIN_Q_STOPPED = 1,
248 NVME_CTRL_STARTED_ONCE = 2,
249 NVME_CTRL_STOPPED = 3,
254 enum nvme_ctrl_state state;
257 struct mutex scan_lock;
258 const struct nvme_ctrl_ops *ops;
259 struct request_queue *admin_q;
260 struct request_queue *connect_q;
261 struct request_queue *fabrics_q;
265 struct blk_mq_tag_set *tagset;
266 struct blk_mq_tag_set *admin_tagset;
267 struct list_head namespaces;
268 struct rw_semaphore namespaces_rwsem;
269 struct device ctrl_device;
270 struct device *device; /* char device */
271 #ifdef CONFIG_NVME_HWMON
272 struct device *hwmon_device;
275 struct work_struct reset_work;
276 struct work_struct delete_work;
277 wait_queue_head_t state_wq;
279 struct nvme_subsystem *subsys;
280 struct list_head subsys_entry;
282 struct opal_dev *opal_dev;
294 u32 max_integrity_segments;
295 u32 max_discard_sectors;
296 u32 max_discard_segments;
297 u32 max_zeroes_sectors;
298 #ifdef CONFIG_BLK_DEV_ZONED
307 atomic_t abort_limit;
319 unsigned int shutdown_timeout;
322 unsigned long quirks;
323 struct nvme_id_power_state psd[32];
324 struct nvme_effects_log *effects;
326 struct work_struct scan_work;
327 struct work_struct async_event_work;
328 struct delayed_work ka_work;
329 struct delayed_work failfast_work;
330 struct nvme_command ka_cmd;
331 struct work_struct fw_act_work;
332 unsigned long events;
334 #ifdef CONFIG_NVME_MULTIPATH
335 /* asymmetric namespace access: */
340 struct mutex ana_lock;
341 struct nvme_ana_rsp_hdr *ana_log_buf;
343 struct timer_list anatt_timer;
344 struct work_struct ana_work;
347 #ifdef CONFIG_NVME_AUTH
348 struct work_struct dhchap_auth_work;
349 struct mutex dhchap_auth_mutex;
350 struct nvme_dhchap_queue_context *dhchap_ctxs;
351 struct nvme_dhchap_key *host_key;
352 struct nvme_dhchap_key *ctrl_key;
356 /* Power saving configuration */
357 u64 ps_max_latency_us;
373 struct nvmf_ctrl_options *opts;
375 struct page *discard_page;
376 unsigned long discard_page_busy;
378 struct nvme_fault_inject fault_inject;
380 enum nvme_ctrl_type cntrltype;
381 enum nvme_dctype dctype;
389 struct nvme_subsystem {
393 * Because we unregister the device on the last put we need
394 * a separate refcount.
397 struct list_head entry;
399 struct list_head ctrls;
400 struct list_head nsheads;
401 char subnqn[NVMF_NQN_SIZE];
404 char firmware_rev[8];
406 enum nvme_subsys_type subtype;
408 u16 awupf; /* 0's based awupf value. */
410 #ifdef CONFIG_NVME_MULTIPATH
411 enum nvme_iopolicy iopolicy;
416 * Container structure for uniqueue namespace identifiers.
426 * Anchor structure for namespaces. There is one for each namespace in a
427 * NVMe subsystem that any of our controllers can see, and the namespace
428 * structure for each controller is chained of it. For private namespaces
429 * there is a 1:1 relation to our namespace structures, that is ->list
430 * only ever has a single entry for private namespaces.
432 struct nvme_ns_head {
433 struct list_head list;
434 struct srcu_struct srcu;
435 struct nvme_subsystem *subsys;
437 struct nvme_ns_ids ids;
438 struct list_head entry;
442 struct nvme_effects_log *effects;
445 struct device cdev_device;
447 struct gendisk *disk;
448 #ifdef CONFIG_NVME_MULTIPATH
449 struct bio_list requeue_list;
450 spinlock_t requeue_lock;
451 struct work_struct requeue_work;
454 #define NVME_NSHEAD_DISK_LIVE 0
455 struct nvme_ns __rcu *current_path[];
459 static inline bool nvme_ns_head_multipath(struct nvme_ns_head *head)
461 return IS_ENABLED(CONFIG_NVME_MULTIPATH) && head->disk;
464 enum nvme_ns_features {
465 NVME_NS_EXT_LBAS = 1 << 0, /* support extended LBA format */
466 NVME_NS_METADATA_SUPPORTED = 1 << 1, /* support getting generated md */
467 NVME_NS_DEAC, /* DEAC bit in Write Zeores supported */
471 struct list_head list;
473 struct nvme_ctrl *ctrl;
474 struct request_queue *queue;
475 struct gendisk *disk;
476 #ifdef CONFIG_NVME_MULTIPATH
477 enum nvme_ana_state ana_state;
480 struct list_head siblings;
482 struct nvme_ns_head *head;
491 #ifdef CONFIG_BLK_DEV_ZONED
494 unsigned long features;
496 #define NVME_NS_REMOVING 0
497 #define NVME_NS_ANA_PENDING 2
498 #define NVME_NS_FORCE_RO 3
499 #define NVME_NS_READY 4
502 struct device cdev_device;
504 struct nvme_fault_inject fault_inject;
508 /* NVMe ns supports metadata actions by the controller (generate/strip) */
509 static inline bool nvme_ns_has_pi(struct nvme_ns *ns)
511 return ns->pi_type && ns->ms == ns->pi_size;
514 struct nvme_ctrl_ops {
516 struct module *module;
518 #define NVME_F_FABRICS (1 << 0)
519 #define NVME_F_METADATA_SUPPORTED (1 << 1)
520 #define NVME_F_BLOCKING (1 << 2)
522 const struct attribute_group **dev_attr_groups;
523 int (*reg_read32)(struct nvme_ctrl *ctrl, u32 off, u32 *val);
524 int (*reg_write32)(struct nvme_ctrl *ctrl, u32 off, u32 val);
525 int (*reg_read64)(struct nvme_ctrl *ctrl, u32 off, u64 *val);
526 void (*free_ctrl)(struct nvme_ctrl *ctrl);
527 void (*submit_async_event)(struct nvme_ctrl *ctrl);
528 void (*delete_ctrl)(struct nvme_ctrl *ctrl);
529 void (*stop_ctrl)(struct nvme_ctrl *ctrl);
530 int (*get_address)(struct nvme_ctrl *ctrl, char *buf, int size);
531 void (*print_device_info)(struct nvme_ctrl *ctrl);
532 bool (*supports_pci_p2pdma)(struct nvme_ctrl *ctrl);
536 * nvme command_id is constructed as such:
537 * | xxxx | xxxxxxxxxxxx |
540 #define nvme_genctr_mask(gen) (gen & 0xf)
541 #define nvme_cid_install_genctr(gen) (nvme_genctr_mask(gen) << 12)
542 #define nvme_genctr_from_cid(cid) ((cid & 0xf000) >> 12)
543 #define nvme_tag_from_cid(cid) (cid & 0xfff)
545 static inline u16 nvme_cid(struct request *rq)
547 return nvme_cid_install_genctr(nvme_req(rq)->genctr) | rq->tag;
550 static inline struct request *nvme_find_rq(struct blk_mq_tags *tags,
553 u8 genctr = nvme_genctr_from_cid(command_id);
554 u16 tag = nvme_tag_from_cid(command_id);
557 rq = blk_mq_tag_to_rq(tags, tag);
559 pr_err("could not locate request for tag %#x\n",
563 if (unlikely(nvme_genctr_mask(nvme_req(rq)->genctr) != genctr)) {
564 dev_err(nvme_req(rq)->ctrl->device,
565 "request %#x genctr mismatch (got %#x expected %#x)\n",
566 tag, genctr, nvme_genctr_mask(nvme_req(rq)->genctr));
572 static inline struct request *nvme_cid_to_rq(struct blk_mq_tags *tags,
575 return blk_mq_tag_to_rq(tags, nvme_tag_from_cid(command_id));
579 * Return the length of the string without the space padding
581 static inline int nvme_strlen(char *s, int len)
583 while (s[len - 1] == ' ')
588 static inline void nvme_print_device_info(struct nvme_ctrl *ctrl)
590 struct nvme_subsystem *subsys = ctrl->subsys;
592 if (ctrl->ops->print_device_info) {
593 ctrl->ops->print_device_info(ctrl);
597 dev_err(ctrl->device,
598 "VID:%04x model:%.*s firmware:%.*s\n", subsys->vendor_id,
599 nvme_strlen(subsys->model, sizeof(subsys->model)),
600 subsys->model, nvme_strlen(subsys->firmware_rev,
601 sizeof(subsys->firmware_rev)),
602 subsys->firmware_rev);
605 #ifdef CONFIG_FAULT_INJECTION_DEBUG_FS
606 void nvme_fault_inject_init(struct nvme_fault_inject *fault_inj,
607 const char *dev_name);
608 void nvme_fault_inject_fini(struct nvme_fault_inject *fault_inject);
609 void nvme_should_fail(struct request *req);
611 static inline void nvme_fault_inject_init(struct nvme_fault_inject *fault_inj,
612 const char *dev_name)
615 static inline void nvme_fault_inject_fini(struct nvme_fault_inject *fault_inj)
618 static inline void nvme_should_fail(struct request *req) {}
621 bool nvme_wait_reset(struct nvme_ctrl *ctrl);
622 int nvme_try_sched_reset(struct nvme_ctrl *ctrl);
624 static inline int nvme_reset_subsystem(struct nvme_ctrl *ctrl)
628 if (!ctrl->subsystem)
630 if (!nvme_wait_reset(ctrl))
633 ret = ctrl->ops->reg_write32(ctrl, NVME_REG_NSSR, 0x4E564D65);
637 return nvme_try_sched_reset(ctrl);
641 * Convert a 512B sector number to a device logical block number.
643 static inline u64 nvme_sect_to_lba(struct nvme_ns *ns, sector_t sector)
645 return sector >> (ns->lba_shift - SECTOR_SHIFT);
649 * Convert a device logical block number to a 512B sector number.
651 static inline sector_t nvme_lba_to_sect(struct nvme_ns *ns, u64 lba)
653 return lba << (ns->lba_shift - SECTOR_SHIFT);
657 * Convert byte length to nvme's 0-based num dwords
659 static inline u32 nvme_bytes_to_numd(size_t len)
661 return (len >> 2) - 1;
664 static inline bool nvme_is_ana_error(u16 status)
666 switch (status & 0x7ff) {
667 case NVME_SC_ANA_TRANSITION:
668 case NVME_SC_ANA_INACCESSIBLE:
669 case NVME_SC_ANA_PERSISTENT_LOSS:
676 static inline bool nvme_is_path_error(u16 status)
678 /* check for a status code type of 'path related status' */
679 return (status & 0x700) == 0x300;
683 * Fill in the status and result information from the CQE, and then figure out
684 * if blk-mq will need to use IPI magic to complete the request, and if yes do
685 * so. If not let the caller complete the request without an indirect function
688 static inline bool nvme_try_complete_req(struct request *req, __le16 status,
689 union nvme_result result)
691 struct nvme_request *rq = nvme_req(req);
692 struct nvme_ctrl *ctrl = rq->ctrl;
694 if (!(ctrl->quirks & NVME_QUIRK_SKIP_CID_GEN))
697 rq->status = le16_to_cpu(status) >> 1;
699 /* inject error when permitted by fault injection framework */
700 nvme_should_fail(req);
701 if (unlikely(blk_should_fake_timeout(req->q)))
703 return blk_mq_complete_request_remote(req);
706 static inline void nvme_get_ctrl(struct nvme_ctrl *ctrl)
708 get_device(ctrl->device);
711 static inline void nvme_put_ctrl(struct nvme_ctrl *ctrl)
713 put_device(ctrl->device);
716 static inline bool nvme_is_aen_req(u16 qid, __u16 command_id)
719 nvme_tag_from_cid(command_id) >= NVME_AQ_BLK_MQ_DEPTH;
722 void nvme_complete_rq(struct request *req);
723 void nvme_complete_batch_req(struct request *req);
725 static __always_inline void nvme_complete_batch(struct io_comp_batch *iob,
726 void (*fn)(struct request *rq))
730 rq_list_for_each(&iob->req_list, req) {
732 nvme_complete_batch_req(req);
734 blk_mq_end_request_batch(iob);
737 blk_status_t nvme_host_path_error(struct request *req);
738 bool nvme_cancel_request(struct request *req, void *data);
739 void nvme_cancel_tagset(struct nvme_ctrl *ctrl);
740 void nvme_cancel_admin_tagset(struct nvme_ctrl *ctrl);
741 bool nvme_change_ctrl_state(struct nvme_ctrl *ctrl,
742 enum nvme_ctrl_state new_state);
743 int nvme_disable_ctrl(struct nvme_ctrl *ctrl, bool shutdown);
744 int nvme_enable_ctrl(struct nvme_ctrl *ctrl);
745 int nvme_init_ctrl(struct nvme_ctrl *ctrl, struct device *dev,
746 const struct nvme_ctrl_ops *ops, unsigned long quirks);
747 void nvme_uninit_ctrl(struct nvme_ctrl *ctrl);
748 void nvme_start_ctrl(struct nvme_ctrl *ctrl);
749 void nvme_stop_ctrl(struct nvme_ctrl *ctrl);
750 int nvme_init_ctrl_finish(struct nvme_ctrl *ctrl, bool was_suspended);
751 int nvme_alloc_admin_tag_set(struct nvme_ctrl *ctrl, struct blk_mq_tag_set *set,
752 const struct blk_mq_ops *ops, unsigned int cmd_size);
753 void nvme_remove_admin_tag_set(struct nvme_ctrl *ctrl);
754 int nvme_alloc_io_tag_set(struct nvme_ctrl *ctrl, struct blk_mq_tag_set *set,
755 const struct blk_mq_ops *ops, unsigned int nr_maps,
756 unsigned int cmd_size);
757 void nvme_remove_io_tag_set(struct nvme_ctrl *ctrl);
759 void nvme_remove_namespaces(struct nvme_ctrl *ctrl);
761 void nvme_complete_async_event(struct nvme_ctrl *ctrl, __le16 status,
762 volatile union nvme_result *res);
764 void nvme_quiesce_io_queues(struct nvme_ctrl *ctrl);
765 void nvme_unquiesce_io_queues(struct nvme_ctrl *ctrl);
766 void nvme_quiesce_admin_queue(struct nvme_ctrl *ctrl);
767 void nvme_unquiesce_admin_queue(struct nvme_ctrl *ctrl);
768 void nvme_mark_namespaces_dead(struct nvme_ctrl *ctrl);
769 void nvme_sync_queues(struct nvme_ctrl *ctrl);
770 void nvme_sync_io_queues(struct nvme_ctrl *ctrl);
771 void nvme_unfreeze(struct nvme_ctrl *ctrl);
772 void nvme_wait_freeze(struct nvme_ctrl *ctrl);
773 int nvme_wait_freeze_timeout(struct nvme_ctrl *ctrl, long timeout);
774 void nvme_start_freeze(struct nvme_ctrl *ctrl);
776 static inline enum req_op nvme_req_op(struct nvme_command *cmd)
778 return nvme_is_write(cmd) ? REQ_OP_DRV_OUT : REQ_OP_DRV_IN;
781 #define NVME_QID_ANY -1
782 void nvme_init_request(struct request *req, struct nvme_command *cmd);
783 void nvme_cleanup_cmd(struct request *req);
784 blk_status_t nvme_setup_cmd(struct nvme_ns *ns, struct request *req);
785 blk_status_t nvme_fail_nonready_command(struct nvme_ctrl *ctrl,
786 struct request *req);
787 bool __nvme_check_ready(struct nvme_ctrl *ctrl, struct request *rq,
790 static inline bool nvme_check_ready(struct nvme_ctrl *ctrl, struct request *rq,
793 if (likely(ctrl->state == NVME_CTRL_LIVE))
795 if (ctrl->ops->flags & NVME_F_FABRICS &&
796 ctrl->state == NVME_CTRL_DELETING)
798 return __nvme_check_ready(ctrl, rq, queue_live);
802 * NSID shall be unique for all shared namespaces, or if at least one of the
803 * following conditions is met:
804 * 1. Namespace Management is supported by the controller
805 * 2. ANA is supported by the controller
806 * 3. NVM Set are supported by the controller
808 * In other case, private namespace are not required to report a unique NSID.
810 static inline bool nvme_is_unique_nsid(struct nvme_ctrl *ctrl,
811 struct nvme_ns_head *head)
813 return head->shared ||
814 (ctrl->oacs & NVME_CTRL_OACS_NS_MNGT_SUPP) ||
815 (ctrl->subsys->cmic & NVME_CTRL_CMIC_ANA) ||
816 (ctrl->ctratt & NVME_CTRL_CTRATT_NVM_SETS);
819 int nvme_submit_sync_cmd(struct request_queue *q, struct nvme_command *cmd,
820 void *buf, unsigned bufflen);
821 int __nvme_submit_sync_cmd(struct request_queue *q, struct nvme_command *cmd,
822 union nvme_result *result, void *buffer, unsigned bufflen,
823 int qid, int at_head,
824 blk_mq_req_flags_t flags);
825 int nvme_set_features(struct nvme_ctrl *dev, unsigned int fid,
826 unsigned int dword11, void *buffer, size_t buflen,
828 int nvme_get_features(struct nvme_ctrl *dev, unsigned int fid,
829 unsigned int dword11, void *buffer, size_t buflen,
831 int nvme_set_queue_count(struct nvme_ctrl *ctrl, int *count);
832 void nvme_stop_keep_alive(struct nvme_ctrl *ctrl);
833 int nvme_reset_ctrl(struct nvme_ctrl *ctrl);
834 int nvme_reset_ctrl_sync(struct nvme_ctrl *ctrl);
835 int nvme_delete_ctrl(struct nvme_ctrl *ctrl);
836 void nvme_queue_scan(struct nvme_ctrl *ctrl);
837 int nvme_get_log(struct nvme_ctrl *ctrl, u32 nsid, u8 log_page, u8 lsp, u8 csi,
838 void *log, size_t size, u64 offset);
839 bool nvme_tryget_ns_head(struct nvme_ns_head *head);
840 void nvme_put_ns_head(struct nvme_ns_head *head);
841 int nvme_cdev_add(struct cdev *cdev, struct device *cdev_device,
842 const struct file_operations *fops, struct module *owner);
843 void nvme_cdev_del(struct cdev *cdev, struct device *cdev_device);
844 int nvme_ioctl(struct block_device *bdev, fmode_t mode,
845 unsigned int cmd, unsigned long arg);
846 long nvme_ns_chr_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
847 int nvme_ns_head_ioctl(struct block_device *bdev, fmode_t mode,
848 unsigned int cmd, unsigned long arg);
849 long nvme_ns_head_chr_ioctl(struct file *file, unsigned int cmd,
851 long nvme_dev_ioctl(struct file *file, unsigned int cmd,
853 int nvme_ns_chr_uring_cmd_iopoll(struct io_uring_cmd *ioucmd,
854 struct io_comp_batch *iob, unsigned int poll_flags);
855 int nvme_ns_head_chr_uring_cmd_iopoll(struct io_uring_cmd *ioucmd,
856 struct io_comp_batch *iob, unsigned int poll_flags);
857 int nvme_ns_chr_uring_cmd(struct io_uring_cmd *ioucmd,
858 unsigned int issue_flags);
859 int nvme_ns_head_chr_uring_cmd(struct io_uring_cmd *ioucmd,
860 unsigned int issue_flags);
861 int nvme_getgeo(struct block_device *bdev, struct hd_geometry *geo);
862 int nvme_dev_uring_cmd(struct io_uring_cmd *ioucmd, unsigned int issue_flags);
864 extern const struct attribute_group *nvme_ns_id_attr_groups[];
865 extern const struct pr_ops nvme_pr_ops;
866 extern const struct block_device_operations nvme_ns_head_ops;
867 extern const struct attribute_group nvme_dev_attrs_group;
869 struct nvme_ns *nvme_find_path(struct nvme_ns_head *head);
870 #ifdef CONFIG_NVME_MULTIPATH
871 static inline bool nvme_ctrl_use_ana(struct nvme_ctrl *ctrl)
873 return ctrl->ana_log_buf != NULL;
876 void nvme_mpath_unfreeze(struct nvme_subsystem *subsys);
877 void nvme_mpath_wait_freeze(struct nvme_subsystem *subsys);
878 void nvme_mpath_start_freeze(struct nvme_subsystem *subsys);
879 void nvme_mpath_default_iopolicy(struct nvme_subsystem *subsys);
880 void nvme_failover_req(struct request *req);
881 void nvme_kick_requeue_lists(struct nvme_ctrl *ctrl);
882 int nvme_mpath_alloc_disk(struct nvme_ctrl *ctrl,struct nvme_ns_head *head);
883 void nvme_mpath_add_disk(struct nvme_ns *ns, __le32 anagrpid);
884 void nvme_mpath_remove_disk(struct nvme_ns_head *head);
885 int nvme_mpath_init_identify(struct nvme_ctrl *ctrl, struct nvme_id_ctrl *id);
886 void nvme_mpath_init_ctrl(struct nvme_ctrl *ctrl);
887 void nvme_mpath_update(struct nvme_ctrl *ctrl);
888 void nvme_mpath_uninit(struct nvme_ctrl *ctrl);
889 void nvme_mpath_stop(struct nvme_ctrl *ctrl);
890 bool nvme_mpath_clear_current_path(struct nvme_ns *ns);
891 void nvme_mpath_revalidate_paths(struct nvme_ns *ns);
892 void nvme_mpath_clear_ctrl_paths(struct nvme_ctrl *ctrl);
893 void nvme_mpath_shutdown_disk(struct nvme_ns_head *head);
894 void nvme_mpath_start_request(struct request *rq);
895 void nvme_mpath_end_request(struct request *rq);
897 static inline void nvme_trace_bio_complete(struct request *req)
899 struct nvme_ns *ns = req->q->queuedata;
901 if ((req->cmd_flags & REQ_NVME_MPATH) && req->bio)
902 trace_block_bio_complete(ns->head->disk->queue, req->bio);
905 extern bool multipath;
906 extern struct device_attribute dev_attr_ana_grpid;
907 extern struct device_attribute dev_attr_ana_state;
908 extern struct device_attribute subsys_attr_iopolicy;
911 #define multipath false
912 static inline bool nvme_ctrl_use_ana(struct nvme_ctrl *ctrl)
916 static inline void nvme_failover_req(struct request *req)
919 static inline void nvme_kick_requeue_lists(struct nvme_ctrl *ctrl)
922 static inline int nvme_mpath_alloc_disk(struct nvme_ctrl *ctrl,
923 struct nvme_ns_head *head)
927 static inline void nvme_mpath_add_disk(struct nvme_ns *ns, __le32 anagrpid)
930 static inline void nvme_mpath_remove_disk(struct nvme_ns_head *head)
933 static inline bool nvme_mpath_clear_current_path(struct nvme_ns *ns)
937 static inline void nvme_mpath_revalidate_paths(struct nvme_ns *ns)
940 static inline void nvme_mpath_clear_ctrl_paths(struct nvme_ctrl *ctrl)
943 static inline void nvme_mpath_shutdown_disk(struct nvme_ns_head *head)
946 static inline void nvme_trace_bio_complete(struct request *req)
949 static inline void nvme_mpath_init_ctrl(struct nvme_ctrl *ctrl)
952 static inline int nvme_mpath_init_identify(struct nvme_ctrl *ctrl,
953 struct nvme_id_ctrl *id)
955 if (ctrl->subsys->cmic & NVME_CTRL_CMIC_ANA)
956 dev_warn(ctrl->device,
957 "Please enable CONFIG_NVME_MULTIPATH for full support of multi-port devices.\n");
960 static inline void nvme_mpath_update(struct nvme_ctrl *ctrl)
963 static inline void nvme_mpath_uninit(struct nvme_ctrl *ctrl)
966 static inline void nvme_mpath_stop(struct nvme_ctrl *ctrl)
969 static inline void nvme_mpath_unfreeze(struct nvme_subsystem *subsys)
972 static inline void nvme_mpath_wait_freeze(struct nvme_subsystem *subsys)
975 static inline void nvme_mpath_start_freeze(struct nvme_subsystem *subsys)
978 static inline void nvme_mpath_default_iopolicy(struct nvme_subsystem *subsys)
981 static inline void nvme_mpath_start_request(struct request *rq)
984 static inline void nvme_mpath_end_request(struct request *rq)
987 #endif /* CONFIG_NVME_MULTIPATH */
989 int nvme_revalidate_zones(struct nvme_ns *ns);
990 int nvme_ns_report_zones(struct nvme_ns *ns, sector_t sector,
991 unsigned int nr_zones, report_zones_cb cb, void *data);
992 #ifdef CONFIG_BLK_DEV_ZONED
993 int nvme_update_zone_info(struct nvme_ns *ns, unsigned lbaf);
994 blk_status_t nvme_setup_zone_mgmt_send(struct nvme_ns *ns, struct request *req,
995 struct nvme_command *cmnd,
996 enum nvme_zone_mgmt_action action);
998 static inline blk_status_t nvme_setup_zone_mgmt_send(struct nvme_ns *ns,
999 struct request *req, struct nvme_command *cmnd,
1000 enum nvme_zone_mgmt_action action)
1002 return BLK_STS_NOTSUPP;
1005 static inline int nvme_update_zone_info(struct nvme_ns *ns, unsigned lbaf)
1007 dev_warn(ns->ctrl->device,
1008 "Please enable CONFIG_BLK_DEV_ZONED to support ZNS devices\n");
1009 return -EPROTONOSUPPORT;
1013 static inline struct nvme_ns *nvme_get_ns_from_dev(struct device *dev)
1015 return dev_to_disk(dev)->private_data;
1018 #ifdef CONFIG_NVME_HWMON
1019 int nvme_hwmon_init(struct nvme_ctrl *ctrl);
1020 void nvme_hwmon_exit(struct nvme_ctrl *ctrl);
1022 static inline int nvme_hwmon_init(struct nvme_ctrl *ctrl)
1027 static inline void nvme_hwmon_exit(struct nvme_ctrl *ctrl)
1032 static inline void nvme_start_request(struct request *rq)
1034 if (rq->cmd_flags & REQ_NVME_MPATH)
1035 nvme_mpath_start_request(rq);
1036 blk_mq_start_request(rq);
1039 static inline bool nvme_ctrl_sgl_supported(struct nvme_ctrl *ctrl)
1041 return ctrl->sgls & ((1 << 0) | (1 << 1));
1044 #ifdef CONFIG_NVME_AUTH
1045 int __init nvme_init_auth(void);
1046 void __exit nvme_exit_auth(void);
1047 int nvme_auth_init_ctrl(struct nvme_ctrl *ctrl);
1048 void nvme_auth_stop(struct nvme_ctrl *ctrl);
1049 int nvme_auth_negotiate(struct nvme_ctrl *ctrl, int qid);
1050 int nvme_auth_wait(struct nvme_ctrl *ctrl, int qid);
1051 void nvme_auth_free(struct nvme_ctrl *ctrl);
1053 static inline int nvme_auth_init_ctrl(struct nvme_ctrl *ctrl)
1057 static inline int __init nvme_init_auth(void)
1061 static inline void __exit nvme_exit_auth(void)
1064 static inline void nvme_auth_stop(struct nvme_ctrl *ctrl) {};
1065 static inline int nvme_auth_negotiate(struct nvme_ctrl *ctrl, int qid)
1067 return -EPROTONOSUPPORT;
1069 static inline int nvme_auth_wait(struct nvme_ctrl *ctrl, int qid)
1071 return NVME_SC_AUTH_REQUIRED;
1073 static inline void nvme_auth_free(struct nvme_ctrl *ctrl) {};
1076 u32 nvme_command_effects(struct nvme_ctrl *ctrl, struct nvme_ns *ns,
1078 u32 nvme_passthru_start(struct nvme_ctrl *ctrl, struct nvme_ns *ns, u8 opcode);
1079 int nvme_execute_rq(struct request *rq, bool at_head);
1080 void nvme_passthru_end(struct nvme_ctrl *ctrl, u32 effects,
1081 struct nvme_command *cmd, int status);
1082 struct nvme_ctrl *nvme_ctrl_from_file(struct file *file);
1083 struct nvme_ns *nvme_find_get_ns(struct nvme_ctrl *ctrl, unsigned nsid);
1084 void nvme_put_ns(struct nvme_ns *ns);
1086 static inline bool nvme_multi_css(struct nvme_ctrl *ctrl)
1088 return (ctrl->ctrl_config & NVME_CC_CSS_MASK) == NVME_CC_CSS_CSI;
1091 #ifdef CONFIG_NVME_VERBOSE_ERRORS
1092 const unsigned char *nvme_get_error_status_str(u16 status);
1093 const unsigned char *nvme_get_opcode_str(u8 opcode);
1094 const unsigned char *nvme_get_admin_opcode_str(u8 opcode);
1095 const unsigned char *nvme_get_fabrics_opcode_str(u8 opcode);
1096 #else /* CONFIG_NVME_VERBOSE_ERRORS */
1097 static inline const unsigned char *nvme_get_error_status_str(u16 status)
1101 static inline const unsigned char *nvme_get_opcode_str(u8 opcode)
1105 static inline const unsigned char *nvme_get_admin_opcode_str(u8 opcode)
1110 static inline const unsigned char *nvme_get_fabrics_opcode_str(u8 opcode)
1112 return "Fabrics Cmd";
1114 #endif /* CONFIG_NVME_VERBOSE_ERRORS */
1116 static inline const unsigned char *nvme_opcode_str(int qid, u8 opcode, u8 fctype)
1118 if (opcode == nvme_fabrics_command)
1119 return nvme_get_fabrics_opcode_str(fctype);
1120 return qid ? nvme_get_opcode_str(opcode) :
1121 nvme_get_admin_opcode_str(opcode);
1123 #endif /* _NVME_H */