drivers/nvme/host/nvme.h

   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/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>
  19
  20 #include <trace/events/block.h>
  21
  22 extern unsigned int nvme_io_timeout;
  23 #define NVME_IO_TIMEOUT (nvme_io_timeout * HZ)
  24
  25 extern unsigned int admin_timeout;
  26 #define NVME_ADMIN_TIMEOUT      (admin_timeout * HZ)
  27
  28 #define NVME_DEFAULT_KATO       5
  29
  30 #ifdef CONFIG_ARCH_NO_SG_CHAIN
  31 #define  NVME_INLINE_SG_CNT  0
  32 #define  NVME_INLINE_METADATA_SG_CNT  0
  33 #else
  34 #define  NVME_INLINE_SG_CNT  2
  35 #define  NVME_INLINE_METADATA_SG_CNT  1
  36 #endif
  37
  38 /*
  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.
  42  */
  43 #define NVME_CTRL_PAGE_SHIFT    12
  44 #define NVME_CTRL_PAGE_SIZE     (1 << NVME_CTRL_PAGE_SHIFT)
  45
  46 extern struct workqueue_struct *nvme_wq;
  47 extern struct workqueue_struct *nvme_reset_wq;
  48 extern struct workqueue_struct *nvme_delete_wq;
  49
  50 /*
  51  * List of workarounds for devices that required behavior not specified in
  52  * the standard.
  53  */
  54 enum nvme_quirks {
  55         /*
  56          * Prefers I/O aligned to a stripe size specified in a vendor
  57          * specific Identify field.
  58          */
  59         NVME_QUIRK_STRIPE_SIZE                  = (1 << 0),
  60
  61         /*
  62          * The controller doesn't handle Identify value others than 0 or 1
  63          * correctly.
  64          */
  65         NVME_QUIRK_IDENTIFY_CNS                 = (1 << 1),
  66
  67         /*
  68          * The controller deterministically returns O's on reads to
  69          * logical blocks that deallocate was called on.
  70          */
  71         NVME_QUIRK_DEALLOCATE_ZEROES            = (1 << 2),
  72
  73         /*
  74          * The controller needs a delay before starts checking the device
  75          * readiness, which is done by reading the NVME_CSTS_RDY bit.
  76          */
  77         NVME_QUIRK_DELAY_BEFORE_CHK_RDY         = (1 << 3),
  78
  79         /*
  80          * APST should not be used.
  81          */
  82         NVME_QUIRK_NO_APST                      = (1 << 4),
  83
  84         /*
  85          * The deepest sleep state should not be used.
  86          */
  87         NVME_QUIRK_NO_DEEPEST_PS                = (1 << 5),
  88
  89         /*
  90          * Set MEDIUM priority on SQ creation
  91          */
  92         NVME_QUIRK_MEDIUM_PRIO_SQ               = (1 << 7),
  93
  94         /*
  95          * Ignore device provided subnqn.
  96          */
  97         NVME_QUIRK_IGNORE_DEV_SUBNQN            = (1 << 8),
  98
  99         /*
 100          * Broken Write Zeroes.
 101          */
 102         NVME_QUIRK_DISABLE_WRITE_ZEROES         = (1 << 9),
 103
 104         /*
 105          * Force simple suspend/resume path.
 106          */
 107         NVME_QUIRK_SIMPLE_SUSPEND               = (1 << 10),
 108
 109         /*
 110          * Use only one interrupt vector for all queues
 111          */
 112         NVME_QUIRK_SINGLE_VECTOR                = (1 << 11),
 113
 114         /*
 115          * Use non-standard 128 bytes SQEs.
 116          */
 117         NVME_QUIRK_128_BYTES_SQES               = (1 << 12),
 118
 119         /*
 120          * Prevent tag overlap between queues
 121          */
 122         NVME_QUIRK_SHARED_TAGS                  = (1 << 13),
 123
 124         /*
 125          * Don't change the value of the temperature threshold feature
 126          */
 127         NVME_QUIRK_NO_TEMP_THRESH_CHANGE        = (1 << 14),
 128
 129         /*
 130          * The controller doesn't handle the Identify Namespace
 131          * Identification Descriptor list subcommand despite claiming
 132          * NVMe 1.3 compliance.
 133          */
 134         NVME_QUIRK_NO_NS_DESC_LIST              = (1 << 15),
 135
 136         /*
 137          * The controller does not properly handle DMA addresses over
 138          * 48 bits.
 139          */
 140         NVME_QUIRK_DMA_ADDRESS_BITS_48          = (1 << 16),
 141
 142         /*
 143          * The controller requires the command_id value be be limited, so skip
 144          * encoding the generation sequence number.
 145          */
 146         NVME_QUIRK_SKIP_CID_GEN                 = (1 << 17),
 147
 148         /*
 149          * Reports garbage in the namespace identifiers (eui64, nguid, uuid).
 150          */
 151         NVME_QUIRK_BOGUS_NID                    = (1 << 18),
 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                      genctr;
 162         u8                      retries;
 163         u8                      flags;
 164         u16                     status;
 165         struct nvme_ctrl        *ctrl;
 166 };
 167
 168 /*
 169  * Mark a bio as coming in through the mpath node.
 170  */
 171 #define REQ_NVME_MPATH          REQ_DRV
 172
 173 enum {
 174         NVME_REQ_CANCELLED              = (1 << 0),
 175         NVME_REQ_USERCMD                = (1 << 1),
 176 };
 177
 178 static inline struct nvme_request *nvme_req(struct request *req)
 179 {
 180         return blk_mq_rq_to_pdu(req);
 181 }
 182
 183 static inline u16 nvme_req_qid(struct request *req)
 184 {
 185         if (!req->q->queuedata)
 186                 return 0;
 187
 188         return req->mq_hctx->queue_num + 1;
 189 }
 190
 191 /* The below value is the specific amount of delay needed before checking
 192  * readiness in case of the PCI_DEVICE(0x1c58, 0x0003), which needs the
 193  * NVME_QUIRK_DELAY_BEFORE_CHK_RDY quirk enabled. The value (in ms) was
 194  * found empirically.
 195  */
 196 #define NVME_QUIRK_DELAY_AMOUNT         2300
 197
 198 /*
 199  * enum nvme_ctrl_state: Controller state
 200  *
 201  * @NVME_CTRL_NEW:              New controller just allocated, initial state
 202  * @NVME_CTRL_LIVE:             Controller is connected and I/O capable
 203  * @NVME_CTRL_RESETTING:        Controller is resetting (or scheduled reset)
 204  * @NVME_CTRL_CONNECTING:       Controller is disconnected, now connecting the
 205  *                              transport
 206  * @NVME_CTRL_DELETING:         Controller is deleting (or scheduled deletion)
 207  * @NVME_CTRL_DELETING_NOIO:    Controller is deleting and I/O is not
 208  *                              disabled/failed immediately. This state comes
 209  *                              after all async event processing took place and
 210  *                              before ns removal and the controller deletion
 211  *                              progress
 212  * @NVME_CTRL_DEAD:             Controller is non-present/unresponsive during
 213  *                              shutdown or removal. In this case we forcibly
 214  *                              kill all inflight I/O as they have no chance to
 215  *                              complete
 216  */
 217 enum nvme_ctrl_state {
 218         NVME_CTRL_NEW,
 219         NVME_CTRL_LIVE,
 220         NVME_CTRL_RESETTING,
 221         NVME_CTRL_CONNECTING,
 222         NVME_CTRL_DELETING,
 223         NVME_CTRL_DELETING_NOIO,
 224         NVME_CTRL_DEAD,
 225 };
 226
 227 struct nvme_fault_inject {
 228 #ifdef CONFIG_FAULT_INJECTION_DEBUG_FS
 229         struct fault_attr attr;
 230         struct dentry *parent;
 231         bool dont_retry;        /* DNR, do not retry */
 232         u16 status;             /* status code */
 233 #endif
 234 };
 235
 236 struct nvme_ctrl {
 237         bool comp_seen;
 238         enum nvme_ctrl_state state;
 239         bool identified;
 240         spinlock_t lock;
 241         struct mutex scan_lock;
 242         const struct nvme_ctrl_ops *ops;
 243         struct request_queue *admin_q;
 244         struct request_queue *connect_q;
 245         struct request_queue *fabrics_q;
 246         struct device *dev;
 247         int instance;
 248         int numa_node;
 249         struct blk_mq_tag_set *tagset;
 250         struct blk_mq_tag_set *admin_tagset;
 251         struct list_head namespaces;
 252         struct rw_semaphore namespaces_rwsem;
 253         struct device ctrl_device;
 254         struct device *device;  /* char device */
 255 #ifdef CONFIG_NVME_HWMON
 256         struct device *hwmon_device;
 257 #endif
 258         struct cdev cdev;
 259         struct work_struct reset_work;
 260         struct work_struct delete_work;
 261         wait_queue_head_t state_wq;
 262
 263         struct nvme_subsystem *subsys;
 264         struct list_head subsys_entry;
 265
 266         struct opal_dev *opal_dev;
 267
 268         char name[12];
 269         u16 cntlid;
 270
 271         u32 ctrl_config;
 272         u16 mtfa;
 273         u32 queue_count;
 274
 275         u64 cap;
 276         u32 max_hw_sectors;
 277         u32 max_segments;
 278         u32 max_integrity_segments;
 279         u32 max_discard_sectors;
 280         u32 max_discard_segments;
 281         u32 max_zeroes_sectors;
 282 #ifdef CONFIG_BLK_DEV_ZONED
 283         u32 max_zone_append;
 284 #endif
 285         u16 crdt[3];
 286         u16 oncs;
 287         u16 oacs;
 288         u16 nssa;
 289         u16 nr_streams;
 290         u16 sqsize;
 291         u32 max_namespaces;
 292         atomic_t abort_limit;
 293         u8 vwc;
 294         u32 vs;
 295         u32 sgls;
 296         u16 kas;
 297         u8 npss;
 298         u8 apsta;
 299         u16 wctemp;
 300         u16 cctemp;
 301         u32 oaes;
 302         u32 aen_result;
 303         u32 ctratt;
 304         unsigned int shutdown_timeout;
 305         unsigned int kato;
 306         bool subsystem;
 307         unsigned long quirks;
 308         struct nvme_id_power_state psd[32];
 309         struct nvme_effects_log *effects;
 310         struct xarray cels;
 311         struct work_struct scan_work;
 312         struct work_struct async_event_work;
 313         struct delayed_work ka_work;
 314         struct delayed_work failfast_work;
 315         struct nvme_command ka_cmd;
 316         struct work_struct fw_act_work;
 317         unsigned long events;
 318
 319 #ifdef CONFIG_NVME_MULTIPATH
 320         /* asymmetric namespace access: */
 321         u8 anacap;
 322         u8 anatt;
 323         u32 anagrpmax;
 324         u32 nanagrpid;
 325         struct mutex ana_lock;
 326         struct nvme_ana_rsp_hdr *ana_log_buf;
 327         size_t ana_log_size;
 328         struct timer_list anatt_timer;
 329         struct work_struct ana_work;
 330 #endif
 331
 332         /* Power saving configuration */
 333         u64 ps_max_latency_us;
 334         bool apst_enabled;
 335
 336         /* PCIe only: */
 337         u32 hmpre;
 338         u32 hmmin;
 339         u32 hmminds;
 340         u16 hmmaxd;
 341
 342         /* Fabrics only */
 343         u32 ioccsz;
 344         u32 iorcsz;
 345         u16 icdoff;
 346         u16 maxcmd;
 347         int nr_reconnects;
 348         unsigned long flags;
 349 #define NVME_CTRL_FAILFAST_EXPIRED      0
 350         struct nvmf_ctrl_options *opts;
 351
 352         struct page *discard_page;
 353         unsigned long discard_page_busy;
 354
 355         struct nvme_fault_inject fault_inject;
 356
 357         enum nvme_ctrl_type cntrltype;
 358         enum nvme_dctype dctype;
 359 };
 360
 361 enum nvme_iopolicy {
 362         NVME_IOPOLICY_NUMA,
 363         NVME_IOPOLICY_RR,
 364 };
 365
 366 struct nvme_subsystem {
 367         int                     instance;
 368         struct device           dev;
 369         /*
 370          * Because we unregister the device on the last put we need
 371          * a separate refcount.
 372          */
 373         struct kref             ref;
 374         struct list_head        entry;
 375         struct mutex            lock;
 376         struct list_head        ctrls;
 377         struct list_head        nsheads;
 378         char                    subnqn[NVMF_NQN_SIZE];
 379         char                    serial[20];
 380         char                    model[40];
 381         char                    firmware_rev[8];
 382         u8                      cmic;
 383         u16                     vendor_id;
 384         u16                     awupf;  /* 0's based awupf value. */
 385         struct ida              ns_ida;
 386 #ifdef CONFIG_NVME_MULTIPATH
 387         enum nvme_iopolicy      iopolicy;
 388 #endif
 389 };
 390
 391 /*
 392  * Container structure for uniqueue namespace identifiers.
 393  */
 394 struct nvme_ns_ids {
 395         u8      eui64[8];
 396         u8      nguid[16];
 397         uuid_t  uuid;
 398         u8      csi;
 399 };
 400
 401 /*
 402  * Anchor structure for namespaces.  There is one for each namespace in a
 403  * NVMe subsystem that any of our controllers can see, and the namespace
 404  * structure for each controller is chained of it.  For private namespaces
 405  * there is a 1:1 relation to our namespace structures, that is ->list
 406  * only ever has a single entry for private namespaces.
 407  */
 408 struct nvme_ns_head {
 409         struct list_head        list;
 410         struct srcu_struct      srcu;
 411         struct nvme_subsystem   *subsys;
 412         unsigned                ns_id;
 413         struct nvme_ns_ids      ids;
 414         struct list_head        entry;
 415         struct kref             ref;
 416         bool                    shared;
 417         int                     instance;
 418         struct nvme_effects_log *effects;
 419
 420         struct cdev             cdev;
 421         struct device           cdev_device;
 422
 423         struct gendisk          *disk;
 424 #ifdef CONFIG_NVME_MULTIPATH
 425         struct bio_list         requeue_list;
 426         spinlock_t              requeue_lock;
 427         struct work_struct      requeue_work;
 428         struct mutex            lock;
 429         unsigned long           flags;
 430 #define NVME_NSHEAD_DISK_LIVE   0
 431         struct nvme_ns __rcu    *current_path[];
 432 #endif
 433 };
 434
 435 static inline bool nvme_ns_head_multipath(struct nvme_ns_head *head)
 436 {
 437         return IS_ENABLED(CONFIG_NVME_MULTIPATH) && head->disk;
 438 }
 439
 440 enum nvme_ns_features {
 441         NVME_NS_EXT_LBAS = 1 << 0, /* support extended LBA format */
 442         NVME_NS_METADATA_SUPPORTED = 1 << 1, /* support getting generated md */
 443 };
 444
 445 struct nvme_ns {
 446         struct list_head list;
 447
 448         struct nvme_ctrl *ctrl;
 449         struct request_queue *queue;
 450         struct gendisk *disk;
 451 #ifdef CONFIG_NVME_MULTIPATH
 452         enum nvme_ana_state ana_state;
 453         u32 ana_grpid;
 454 #endif
 455         struct list_head siblings;
 456         struct kref kref;
 457         struct nvme_ns_head *head;
 458
 459         int lba_shift;
 460         u16 ms;
 461         u16 sgs;
 462         u32 sws;
 463         u8 pi_type;
 464 #ifdef CONFIG_BLK_DEV_ZONED
 465         u64 zsze;
 466 #endif
 467         unsigned long features;
 468         unsigned long flags;
 469 #define NVME_NS_REMOVING        0
 470 #define NVME_NS_DEAD            1
 471 #define NVME_NS_ANA_PENDING     2
 472 #define NVME_NS_FORCE_RO        3
 473 #define NVME_NS_READY           4
 474
 475         struct cdev             cdev;
 476         struct device           cdev_device;
 477
 478         struct nvme_fault_inject fault_inject;
 479
 480 };
 481
 482 /* NVMe ns supports metadata actions by the controller (generate/strip) */
 483 static inline bool nvme_ns_has_pi(struct nvme_ns *ns)
 484 {
 485         return ns->pi_type && ns->ms == sizeof(struct t10_pi_tuple);
 486 }
 487
 488 struct nvme_ctrl_ops {
 489         const char *name;
 490         struct module *module;
 491         unsigned int flags;
 492 #define NVME_F_FABRICS                  (1 << 0)
 493 #define NVME_F_METADATA_SUPPORTED       (1 << 1)
 494 #define NVME_F_PCI_P2PDMA               (1 << 2)
 495         int (*reg_read32)(struct nvme_ctrl *ctrl, u32 off, u32 *val);
 496         int (*reg_write32)(struct nvme_ctrl *ctrl, u32 off, u32 val);
 497         int (*reg_read64)(struct nvme_ctrl *ctrl, u32 off, u64 *val);
 498         void (*free_ctrl)(struct nvme_ctrl *ctrl);
 499         void (*submit_async_event)(struct nvme_ctrl *ctrl);
 500         void (*delete_ctrl)(struct nvme_ctrl *ctrl);
 501         int (*get_address)(struct nvme_ctrl *ctrl, char *buf, int size);
 502 };
 503
 504 /*
 505  * nvme command_id is constructed as such:
 506  * | xxxx | xxxxxxxxxxxx |
 507  *   gen    request tag
 508  */
 509 #define nvme_genctr_mask(gen)                   (gen & 0xf)
 510 #define nvme_cid_install_genctr(gen)            (nvme_genctr_mask(gen) << 12)
 511 #define nvme_genctr_from_cid(cid)               ((cid & 0xf000) >> 12)
 512 #define nvme_tag_from_cid(cid)                  (cid & 0xfff)
 513
 514 static inline u16 nvme_cid(struct request *rq)
 515 {
 516         return nvme_cid_install_genctr(nvme_req(rq)->genctr) | rq->tag;
 517 }
 518
 519 static inline struct request *nvme_find_rq(struct blk_mq_tags *tags,
 520                 u16 command_id)
 521 {
 522         u8 genctr = nvme_genctr_from_cid(command_id);
 523         u16 tag = nvme_tag_from_cid(command_id);
 524         struct request *rq;
 525
 526         rq = blk_mq_tag_to_rq(tags, tag);
 527         if (unlikely(!rq)) {
 528                 pr_err("could not locate request for tag %#x\n",
 529                         tag);
 530                 return NULL;
 531         }
 532         if (unlikely(nvme_genctr_mask(nvme_req(rq)->genctr) != genctr)) {
 533                 dev_err(nvme_req(rq)->ctrl->device,
 534                         "request %#x genctr mismatch (got %#x expected %#x)\n",
 535                         tag, genctr, nvme_genctr_mask(nvme_req(rq)->genctr));
 536                 return NULL;
 537         }
 538         return rq;
 539 }
 540
 541 static inline struct request *nvme_cid_to_rq(struct blk_mq_tags *tags,
 542                 u16 command_id)
 543 {
 544         return blk_mq_tag_to_rq(tags, nvme_tag_from_cid(command_id));
 545 }
 546
 547 #ifdef CONFIG_FAULT_INJECTION_DEBUG_FS
 548 void nvme_fault_inject_init(struct nvme_fault_inject *fault_inj,
 549                             const char *dev_name);
 550 void nvme_fault_inject_fini(struct nvme_fault_inject *fault_inject);
 551 void nvme_should_fail(struct request *req);
 552 #else
 553 static inline void nvme_fault_inject_init(struct nvme_fault_inject *fault_inj,
 554                                           const char *dev_name)
 555 {
 556 }
 557 static inline void nvme_fault_inject_fini(struct nvme_fault_inject *fault_inj)
 558 {
 559 }
 560 static inline void nvme_should_fail(struct request *req) {}
 561 #endif
 562
 563 static inline int nvme_reset_subsystem(struct nvme_ctrl *ctrl)
 564 {
 565         if (!ctrl->subsystem)
 566                 return -ENOTTY;
 567         return ctrl->ops->reg_write32(ctrl, NVME_REG_NSSR, 0x4E564D65);
 568 }
 569
 570 /*
 571  * Convert a 512B sector number to a device logical block number.
 572  */
 573 static inline u64 nvme_sect_to_lba(struct nvme_ns *ns, sector_t sector)
 574 {
 575         return sector >> (ns->lba_shift - SECTOR_SHIFT);
 576 }
 577
 578 /*
 579  * Convert a device logical block number to a 512B sector number.
 580  */
 581 static inline sector_t nvme_lba_to_sect(struct nvme_ns *ns, u64 lba)
 582 {
 583         return lba << (ns->lba_shift - SECTOR_SHIFT);
 584 }
 585
 586 /*
 587  * Convert byte length to nvme's 0-based num dwords
 588  */
 589 static inline u32 nvme_bytes_to_numd(size_t len)
 590 {
 591         return (len >> 2) - 1;
 592 }
 593
 594 static inline bool nvme_is_ana_error(u16 status)
 595 {
 596         switch (status & 0x7ff) {
 597         case NVME_SC_ANA_TRANSITION:
 598         case NVME_SC_ANA_INACCESSIBLE:
 599         case NVME_SC_ANA_PERSISTENT_LOSS:
 600                 return true;
 601         default:
 602                 return false;
 603         }
 604 }
 605
 606 static inline bool nvme_is_path_error(u16 status)
 607 {
 608         /* check for a status code type of 'path related status' */
 609         return (status & 0x700) == 0x300;
 610 }
 611
 612 /*
 613  * Fill in the status and result information from the CQE, and then figure out
 614  * if blk-mq will need to use IPI magic to complete the request, and if yes do
 615  * so.  If not let the caller complete the request without an indirect function
 616  * call.
 617  */
 618 static inline bool nvme_try_complete_req(struct request *req, __le16 status,
 619                 union nvme_result result)
 620 {
 621         struct nvme_request *rq = nvme_req(req);
 622
 623         rq->status = le16_to_cpu(status) >> 1;
 624         rq->result = result;
 625         /* inject error when permitted by fault injection framework */
 626         nvme_should_fail(req);
 627         if (unlikely(blk_should_fake_timeout(req->q)))
 628                 return true;
 629         return blk_mq_complete_request_remote(req);
 630 }
 631
 632 static inline void nvme_get_ctrl(struct nvme_ctrl *ctrl)
 633 {
 634         get_device(ctrl->device);
 635 }
 636
 637 static inline void nvme_put_ctrl(struct nvme_ctrl *ctrl)
 638 {
 639         put_device(ctrl->device);
 640 }
 641
 642 static inline bool nvme_is_aen_req(u16 qid, __u16 command_id)
 643 {
 644         return !qid &&
 645                 nvme_tag_from_cid(command_id) >= NVME_AQ_BLK_MQ_DEPTH;
 646 }
 647
 648 void nvme_complete_rq(struct request *req);
 649 blk_status_t nvme_host_path_error(struct request *req);
 650 bool nvme_cancel_request(struct request *req, void *data, bool reserved);
 651 void nvme_cancel_tagset(struct nvme_ctrl *ctrl);
 652 void nvme_cancel_admin_tagset(struct nvme_ctrl *ctrl);
 653 bool nvme_change_ctrl_state(struct nvme_ctrl *ctrl,
 654                 enum nvme_ctrl_state new_state);
 655 bool nvme_wait_reset(struct nvme_ctrl *ctrl);
 656 int nvme_disable_ctrl(struct nvme_ctrl *ctrl);
 657 int nvme_enable_ctrl(struct nvme_ctrl *ctrl);
 658 int nvme_shutdown_ctrl(struct nvme_ctrl *ctrl);
 659 int nvme_init_ctrl(struct nvme_ctrl *ctrl, struct device *dev,
 660                 const struct nvme_ctrl_ops *ops, unsigned long quirks);
 661 void nvme_uninit_ctrl(struct nvme_ctrl *ctrl);
 662 void nvme_start_ctrl(struct nvme_ctrl *ctrl);
 663 void nvme_stop_ctrl(struct nvme_ctrl *ctrl);
 664 int nvme_init_ctrl_finish(struct nvme_ctrl *ctrl);
 665
 666 void nvme_remove_namespaces(struct nvme_ctrl *ctrl);
 667
 668 int nvme_sec_submit(void *data, u16 spsp, u8 secp, void *buffer, size_t len,
 669                 bool send);
 670
 671 void nvme_complete_async_event(struct nvme_ctrl *ctrl, __le16 status,
 672                 volatile union nvme_result *res);
 673
 674 void nvme_stop_queues(struct nvme_ctrl *ctrl);
 675 void nvme_start_queues(struct nvme_ctrl *ctrl);
 676 void nvme_kill_queues(struct nvme_ctrl *ctrl);
 677 void nvme_sync_queues(struct nvme_ctrl *ctrl);
 678 void nvme_sync_io_queues(struct nvme_ctrl *ctrl);
 679 void nvme_unfreeze(struct nvme_ctrl *ctrl);
 680 void nvme_wait_freeze(struct nvme_ctrl *ctrl);
 681 int nvme_wait_freeze_timeout(struct nvme_ctrl *ctrl, long timeout);
 682 void nvme_start_freeze(struct nvme_ctrl *ctrl);
 683
 684 #define NVME_QID_ANY -1
 685 struct request *nvme_alloc_request(struct request_queue *q,
 686                 struct nvme_command *cmd, blk_mq_req_flags_t flags);
 687 void nvme_cleanup_cmd(struct request *req);
 688 blk_status_t nvme_setup_cmd(struct nvme_ns *ns, struct request *req);
 689 blk_status_t nvme_fail_nonready_command(struct nvme_ctrl *ctrl,
 690                 struct request *req);
 691 bool __nvme_check_ready(struct nvme_ctrl *ctrl, struct request *rq,
 692                 bool queue_live);
 693
 694 static inline bool nvme_check_ready(struct nvme_ctrl *ctrl, struct request *rq,
 695                 bool queue_live)
 696 {
 697         if (likely(ctrl->state == NVME_CTRL_LIVE))
 698                 return true;
 699         if (ctrl->ops->flags & NVME_F_FABRICS &&
 700             ctrl->state == NVME_CTRL_DELETING)
 701                 return true;
 702         return __nvme_check_ready(ctrl, rq, queue_live);
 703 }
 704
 705 /*
 706  * NSID shall be unique for all shared namespaces, or if at least one of the
 707  * following conditions is met:
 708  *   1. Namespace Management is supported by the controller
 709  *   2. ANA is supported by the controller
 710  *   3. NVM Set are supported by the controller
 711  *
 712  * In other case, private namespace are not required to report a unique NSID.
 713  */
 714 static inline bool nvme_is_unique_nsid(struct nvme_ctrl *ctrl,
 715                 struct nvme_ns_head *head)
 716 {
 717         return head->shared ||
 718                 (ctrl->oacs & NVME_CTRL_OACS_NS_MNGT_SUPP) ||
 719                 (ctrl->subsys->cmic & NVME_CTRL_CMIC_ANA) ||
 720                 (ctrl->ctratt & NVME_CTRL_CTRATT_NVM_SETS);
 721 }
 722
 723 int nvme_submit_sync_cmd(struct request_queue *q, struct nvme_command *cmd,
 724                 void *buf, unsigned bufflen);
 725 int __nvme_submit_sync_cmd(struct request_queue *q, struct nvme_command *cmd,
 726                 union nvme_result *result, void *buffer, unsigned bufflen,
 727                 unsigned timeout, int qid, int at_head,
 728                 blk_mq_req_flags_t flags);
 729 int nvme_set_features(struct nvme_ctrl *dev, unsigned int fid,
 730                       unsigned int dword11, void *buffer, size_t buflen,
 731                       u32 *result);
 732 int nvme_get_features(struct nvme_ctrl *dev, unsigned int fid,
 733                       unsigned int dword11, void *buffer, size_t buflen,
 734                       u32 *result);
 735 int nvme_set_queue_count(struct nvme_ctrl *ctrl, int *count);
 736 void nvme_stop_keep_alive(struct nvme_ctrl *ctrl);
 737 int nvme_reset_ctrl(struct nvme_ctrl *ctrl);
 738 int nvme_reset_ctrl_sync(struct nvme_ctrl *ctrl);
 739 int nvme_try_sched_reset(struct nvme_ctrl *ctrl);
 740 int nvme_delete_ctrl(struct nvme_ctrl *ctrl);
 741 void nvme_queue_scan(struct nvme_ctrl *ctrl);
 742 int nvme_get_log(struct nvme_ctrl *ctrl, u32 nsid, u8 log_page, u8 lsp, u8 csi,
 743                 void *log, size_t size, u64 offset);
 744 bool nvme_tryget_ns_head(struct nvme_ns_head *head);
 745 void nvme_put_ns_head(struct nvme_ns_head *head);
 746 int nvme_cdev_add(struct cdev *cdev, struct device *cdev_device,
 747                 const struct file_operations *fops, struct module *owner);
 748 void nvme_cdev_del(struct cdev *cdev, struct device *cdev_device);
 749 int nvme_ioctl(struct block_device *bdev, fmode_t mode,
 750                 unsigned int cmd, unsigned long arg);
 751 long nvme_ns_chr_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
 752 int nvme_ns_head_ioctl(struct block_device *bdev, fmode_t mode,
 753                 unsigned int cmd, unsigned long arg);
 754 long nvme_ns_head_chr_ioctl(struct file *file, unsigned int cmd,
 755                 unsigned long arg);
 756 long nvme_dev_ioctl(struct file *file, unsigned int cmd,
 757                 unsigned long arg);
 758 int nvme_getgeo(struct block_device *bdev, struct hd_geometry *geo);
 759
 760 extern const struct attribute_group *nvme_ns_id_attr_groups[];
 761 extern const struct pr_ops nvme_pr_ops;
 762 extern const struct block_device_operations nvme_ns_head_ops;
 763
 764 struct nvme_ns *nvme_find_path(struct nvme_ns_head *head);
 765 #ifdef CONFIG_NVME_MULTIPATH
 766 static inline bool nvme_ctrl_use_ana(struct nvme_ctrl *ctrl)
 767 {
 768         return ctrl->ana_log_buf != NULL;
 769 }
 770
 771 void nvme_mpath_unfreeze(struct nvme_subsystem *subsys);
 772 void nvme_mpath_wait_freeze(struct nvme_subsystem *subsys);
 773 void nvme_mpath_start_freeze(struct nvme_subsystem *subsys);
 774 bool nvme_mpath_set_disk_name(struct nvme_ns *ns, char *disk_name, int *flags);
 775 void nvme_failover_req(struct request *req);
 776 void nvme_kick_requeue_lists(struct nvme_ctrl *ctrl);
 777 int nvme_mpath_alloc_disk(struct nvme_ctrl *ctrl,struct nvme_ns_head *head);
 778 void nvme_mpath_add_disk(struct nvme_ns *ns, struct nvme_id_ns *id);
 779 void nvme_mpath_remove_disk(struct nvme_ns_head *head);
 780 int nvme_mpath_init_identify(struct nvme_ctrl *ctrl, struct nvme_id_ctrl *id);
 781 void nvme_mpath_init_ctrl(struct nvme_ctrl *ctrl);
 782 void nvme_mpath_uninit(struct nvme_ctrl *ctrl);
 783 void nvme_mpath_stop(struct nvme_ctrl *ctrl);
 784 bool nvme_mpath_clear_current_path(struct nvme_ns *ns);
 785 void nvme_mpath_revalidate_paths(struct nvme_ns *ns);
 786 void nvme_mpath_clear_ctrl_paths(struct nvme_ctrl *ctrl);
 787 void nvme_mpath_shutdown_disk(struct nvme_ns_head *head);
 788
 789 static inline void nvme_trace_bio_complete(struct request *req)
 790 {
 791         struct nvme_ns *ns = req->q->queuedata;
 792
 793         if (req->cmd_flags & REQ_NVME_MPATH)
 794                 trace_block_bio_complete(ns->head->disk->queue, req->bio);
 795 }
 796
 797 extern struct device_attribute dev_attr_ana_grpid;
 798 extern struct device_attribute dev_attr_ana_state;
 799 extern struct device_attribute subsys_attr_iopolicy;
 800
 801 #else
 802 static inline bool nvme_ctrl_use_ana(struct nvme_ctrl *ctrl)
 803 {
 804         return false;
 805 }
 806 static inline bool nvme_mpath_set_disk_name(struct nvme_ns *ns, char *disk_name,
 807                 int *flags)
 808 {
 809         return false;
 810 }
 811 static inline void nvme_failover_req(struct request *req)
 812 {
 813 }
 814 static inline void nvme_kick_requeue_lists(struct nvme_ctrl *ctrl)
 815 {
 816 }
 817 static inline int nvme_mpath_alloc_disk(struct nvme_ctrl *ctrl,
 818                 struct nvme_ns_head *head)
 819 {
 820         return 0;
 821 }
 822 static inline void nvme_mpath_add_disk(struct nvme_ns *ns,
 823                 struct nvme_id_ns *id)
 824 {
 825 }
 826 static inline void nvme_mpath_remove_disk(struct nvme_ns_head *head)
 827 {
 828 }
 829 static inline bool nvme_mpath_clear_current_path(struct nvme_ns *ns)
 830 {
 831         return false;
 832 }
 833 static inline void nvme_mpath_revalidate_paths(struct nvme_ns *ns)
 834 {
 835 }
 836 static inline void nvme_mpath_clear_ctrl_paths(struct nvme_ctrl *ctrl)
 837 {
 838 }
 839 static inline void nvme_mpath_shutdown_disk(struct nvme_ns_head *head)
 840 {
 841 }
 842 static inline void nvme_trace_bio_complete(struct request *req)
 843 {
 844 }
 845 static inline void nvme_mpath_init_ctrl(struct nvme_ctrl *ctrl)
 846 {
 847 }
 848 static inline int nvme_mpath_init_identify(struct nvme_ctrl *ctrl,
 849                 struct nvme_id_ctrl *id)
 850 {
 851         if (ctrl->subsys->cmic & NVME_CTRL_CMIC_ANA)
 852                 dev_warn(ctrl->device,
 853 "Please enable CONFIG_NVME_MULTIPATH for full support of multi-port devices.\n");
 854         return 0;
 855 }
 856 static inline void nvme_mpath_uninit(struct nvme_ctrl *ctrl)
 857 {
 858 }
 859 static inline void nvme_mpath_stop(struct nvme_ctrl *ctrl)
 860 {
 861 }
 862 static inline void nvme_mpath_unfreeze(struct nvme_subsystem *subsys)
 863 {
 864 }
 865 static inline void nvme_mpath_wait_freeze(struct nvme_subsystem *subsys)
 866 {
 867 }
 868 static inline void nvme_mpath_start_freeze(struct nvme_subsystem *subsys)
 869 {
 870 }
 871 #endif /* CONFIG_NVME_MULTIPATH */
 872
 873 int nvme_revalidate_zones(struct nvme_ns *ns);
 874 int nvme_ns_report_zones(struct nvme_ns *ns, sector_t sector,
 875                 unsigned int nr_zones, report_zones_cb cb, void *data);
 876 #ifdef CONFIG_BLK_DEV_ZONED
 877 int nvme_update_zone_info(struct nvme_ns *ns, unsigned lbaf);
 878 blk_status_t nvme_setup_zone_mgmt_send(struct nvme_ns *ns, struct request *req,
 879                                        struct nvme_command *cmnd,
 880                                        enum nvme_zone_mgmt_action action);
 881 #else
 882 static inline blk_status_t nvme_setup_zone_mgmt_send(struct nvme_ns *ns,
 883                 struct request *req, struct nvme_command *cmnd,
 884                 enum nvme_zone_mgmt_action action)
 885 {
 886         return BLK_STS_NOTSUPP;
 887 }
 888
 889 static inline int nvme_update_zone_info(struct nvme_ns *ns, unsigned lbaf)
 890 {
 891         dev_warn(ns->ctrl->device,
 892                  "Please enable CONFIG_BLK_DEV_ZONED to support ZNS devices\n");
 893         return -EPROTONOSUPPORT;
 894 }
 895 #endif
 896
 897 static inline struct nvme_ns *nvme_get_ns_from_dev(struct device *dev)
 898 {
 899         return dev_to_disk(dev)->private_data;
 900 }
 901
 902 #ifdef CONFIG_NVME_HWMON
 903 int nvme_hwmon_init(struct nvme_ctrl *ctrl);
 904 void nvme_hwmon_exit(struct nvme_ctrl *ctrl);
 905 #else
 906 static inline int nvme_hwmon_init(struct nvme_ctrl *ctrl)
 907 {
 908         return 0;
 909 }
 910
 911 static inline void nvme_hwmon_exit(struct nvme_ctrl *ctrl)
 912 {
 913 }
 914 #endif
 915
 916 static inline bool nvme_ctrl_sgl_supported(struct nvme_ctrl *ctrl)
 917 {
 918         return ctrl->sgls & ((1 << 0) | (1 << 1));
 919 }
 920
 921 u32 nvme_command_effects(struct nvme_ctrl *ctrl, struct nvme_ns *ns,
 922                          u8 opcode);
 923 int nvme_execute_passthru_rq(struct request *rq);
 924 struct nvme_ctrl *nvme_ctrl_from_file(struct file *file);
 925 struct nvme_ns *nvme_find_get_ns(struct nvme_ctrl *ctrl, unsigned nsid);
 926 void nvme_put_ns(struct nvme_ns *ns);
 927
 928 static inline bool nvme_multi_css(struct nvme_ctrl *ctrl)
 929 {
 930         return (ctrl->ctrl_config & NVME_CC_CSS_MASK) == NVME_CC_CSS_CSI;
 931 }
 932
 933 #endif /* _NVME_H */