fs/fuse/fuse_i.h

   1 /*
   2   FUSE: Filesystem in Userspace
   3   Copyright (C) 2001-2008  Miklos Szeredi <miklos@szeredi.hu>
   4
   5   This program can be distributed under the terms of the GNU GPL.
   6   See the file COPYING.
   7 */
   8
   9 #ifndef _FS_FUSE_I_H
  10 #define _FS_FUSE_I_H
  11
  12 #ifndef pr_fmt
  13 # define pr_fmt(fmt) "fuse: " fmt
  14 #endif
  15
  16 #include <linux/fuse.h>
  17 #include <linux/fs.h>
  18 #include <linux/mount.h>
  19 #include <linux/wait.h>
  20 #include <linux/list.h>
  21 #include <linux/spinlock.h>
  22 #include <linux/mm.h>
  23 #include <linux/backing-dev.h>
  24 #include <linux/mutex.h>
  25 #include <linux/rwsem.h>
  26 #include <linux/rbtree.h>
  27 #include <linux/poll.h>
  28 #include <linux/workqueue.h>
  29 #include <linux/kref.h>
  30 #include <linux/xattr.h>
  31 #include <linux/pid_namespace.h>
  32 #include <linux/refcount.h>
  33 #include <linux/user_namespace.h>
  34
  35 /** Default max number of pages that can be used in a single read request */
  36 #define FUSE_DEFAULT_MAX_PAGES_PER_REQ 32
  37
  38 /** Maximum of max_pages received in init_out */
  39 #define FUSE_MAX_MAX_PAGES 256
  40
  41 /** Bias for fi->writectr, meaning new writepages must not be sent */
  42 #define FUSE_NOWRITE INT_MIN
  43
  44 /** It could be as large as PATH_MAX, but would that have any uses? */
  45 #define FUSE_NAME_MAX 1024
  46
  47 /** Number of dentries for each connection in the control filesystem */
  48 #define FUSE_CTL_NUM_DENTRIES 5
  49
  50 /** List of active connections */
  51 extern struct list_head fuse_conn_list;
  52
  53 /** Global mutex protecting fuse_conn_list and the control filesystem */
  54 extern struct mutex fuse_mutex;
  55
  56 /** Module parameters */
  57 extern unsigned max_user_bgreq;
  58 extern unsigned max_user_congthresh;
  59
  60 /* One forget request */
  61 struct fuse_forget_link {
  62         struct fuse_forget_one forget_one;
  63         struct fuse_forget_link *next;
  64 };
  65
  66 /* Submount lookup tracking */
  67 struct fuse_submount_lookup {
  68         /** Refcount */
  69         refcount_t count;
  70
  71         /** Unique ID, which identifies the inode between userspace
  72          * and kernel */
  73         u64 nodeid;
  74
  75         /** The request used for sending the FORGET message */
  76         struct fuse_forget_link *forget;
  77 };
  78
  79 /** FUSE inode */
  80 struct fuse_inode {
  81         /** Inode data */
  82         struct inode inode;
  83
  84         /** Unique ID, which identifies the inode between userspace
  85          * and kernel */
  86         u64 nodeid;
  87
  88         /** Number of lookups on this inode */
  89         u64 nlookup;
  90
  91         /** The request used for sending the FORGET message */
  92         struct fuse_forget_link *forget;
  93
  94         /** Time in jiffies until the file attributes are valid */
  95         u64 i_time;
  96
  97         /* Which attributes are invalid */
  98         u32 inval_mask;
  99
 100         /** The sticky bit in inode->i_mode may have been removed, so
 101             preserve the original mode */
 102         umode_t orig_i_mode;
 103
 104         /* Cache birthtime */
 105         struct timespec64 i_btime;
 106
 107         /** 64 bit inode number */
 108         u64 orig_ino;
 109
 110         /** Version of last attribute change */
 111         u64 attr_version;
 112
 113         union {
 114                 /* read/write io cache (regular file only) */
 115                 struct {
 116                         /* Files usable in writepage.  Protected by fi->lock */
 117                         struct list_head write_files;
 118
 119                         /* Writepages pending on truncate or fsync */
 120                         struct list_head queued_writes;
 121
 122                         /* Number of sent writes, a negative bias
 123                          * (FUSE_NOWRITE) means more writes are blocked */
 124                         int writectr;
 125
 126                         /** Number of files/maps using page cache */
 127                         int iocachectr;
 128
 129                         /* Waitq for writepage completion */
 130                         wait_queue_head_t page_waitq;
 131
 132                         /* waitq for direct-io completion */
 133                         wait_queue_head_t direct_io_waitq;
 134
 135                         /* List of writepage requestst (pending or sent) */
 136                         struct rb_root writepages;
 137                 };
 138
 139                 /* readdir cache (directory only) */
 140                 struct {
 141                         /* true if fully cached */
 142                         bool cached;
 143
 144                         /* size of cache */
 145                         loff_t size;
 146
 147                         /* position at end of cache (position of next entry) */
 148                         loff_t pos;
 149
 150                         /* version of the cache */
 151                         u64 version;
 152
 153                         /* modification time of directory when cache was
 154                          * started */
 155                         struct timespec64 mtime;
 156
 157                         /* iversion of directory when cache was started */
 158                         u64 iversion;
 159
 160                         /* protects above fields */
 161                         spinlock_t lock;
 162                 } rdc;
 163         };
 164
 165         /** Miscellaneous bits describing inode state */
 166         unsigned long state;
 167
 168         /** Lock for serializing lookup and readdir for back compatibility*/
 169         struct mutex mutex;
 170
 171         /** Lock to protect write related fields */
 172         spinlock_t lock;
 173
 174 #ifdef CONFIG_FUSE_DAX
 175         /*
 176          * Dax specific inode data
 177          */
 178         struct fuse_inode_dax *dax;
 179 #endif
 180         /** Submount specific lookup tracking */
 181         struct fuse_submount_lookup *submount_lookup;
 182 };
 183
 184 /** FUSE inode state bits */
 185 enum {
 186         /** Advise readdirplus  */
 187         FUSE_I_ADVISE_RDPLUS,
 188         /** Initialized with readdirplus */
 189         FUSE_I_INIT_RDPLUS,
 190         /** An operation changing file size is in progress  */
 191         FUSE_I_SIZE_UNSTABLE,
 192         /* Bad inode */
 193         FUSE_I_BAD,
 194         /* Has btime */
 195         FUSE_I_BTIME,
 196         /* Wants or already has page cache IO */
 197         FUSE_I_CACHE_IO_MODE,
 198 };
 199
 200 struct fuse_conn;
 201 struct fuse_mount;
 202 struct fuse_release_args;
 203
 204 /** FUSE specific file data */
 205 struct fuse_file {
 206         /** Fuse connection for this file */
 207         struct fuse_mount *fm;
 208
 209         /* Argument space reserved for release */
 210         struct fuse_release_args *release_args;
 211
 212         /** Kernel file handle guaranteed to be unique */
 213         u64 kh;
 214
 215         /** File handle used by userspace */
 216         u64 fh;
 217
 218         /** Node id of this file */
 219         u64 nodeid;
 220
 221         /** Refcount */
 222         refcount_t count;
 223
 224         /** FOPEN_* flags returned by open */
 225         u32 open_flags;
 226
 227         /** Entry on inode's write_files list */
 228         struct list_head write_entry;
 229
 230         /* Readdir related */
 231         struct {
 232                 /*
 233                  * Protects below fields against (crazy) parallel readdir on
 234                  * same open file.  Uncontended in the normal case.
 235                  */
 236                 struct mutex lock;
 237
 238                 /* Dir stream position */
 239                 loff_t pos;
 240
 241                 /* Offset in cache */
 242                 loff_t cache_off;
 243
 244                 /* Version of cache we are reading */
 245                 u64 version;
 246
 247         } readdir;
 248
 249         /** RB node to be linked on fuse_conn->polled_files */
 250         struct rb_node polled_node;
 251
 252         /** Wait queue head for poll */
 253         wait_queue_head_t poll_wait;
 254
 255         /** Does file hold a fi->iocachectr refcount? */
 256         enum { IOM_NONE, IOM_CACHED, IOM_UNCACHED } iomode;
 257
 258         /** Has flock been performed on this file? */
 259         bool flock:1;
 260 };
 261
 262 /** One input argument of a request */
 263 struct fuse_in_arg {
 264         unsigned size;
 265         const void *value;
 266 };
 267
 268 /** One output argument of a request */
 269 struct fuse_arg {
 270         unsigned size;
 271         void *value;
 272 };
 273
 274 /** FUSE page descriptor */
 275 struct fuse_page_desc {
 276         unsigned int length;
 277         unsigned int offset;
 278 };
 279
 280 struct fuse_args {
 281         uint64_t nodeid;
 282         uint32_t opcode;
 283         uint8_t in_numargs;
 284         uint8_t out_numargs;
 285         uint8_t ext_idx;
 286         bool force:1;
 287         bool noreply:1;
 288         bool nocreds:1;
 289         bool in_pages:1;
 290         bool out_pages:1;
 291         bool user_pages:1;
 292         bool out_argvar:1;
 293         bool page_zeroing:1;
 294         bool page_replace:1;
 295         bool may_block:1;
 296         bool is_ext:1;
 297         struct fuse_in_arg in_args[3];
 298         struct fuse_arg out_args[2];
 299         void (*end)(struct fuse_mount *fm, struct fuse_args *args, int error);
 300 };
 301
 302 struct fuse_args_pages {
 303         struct fuse_args args;
 304         struct page **pages;
 305         struct fuse_page_desc *descs;
 306         unsigned int num_pages;
 307 };
 308
 309 #define FUSE_ARGS(args) struct fuse_args args = {}
 310
 311 /** The request IO state (for asynchronous processing) */
 312 struct fuse_io_priv {
 313         struct kref refcnt;
 314         int async;
 315         spinlock_t lock;
 316         unsigned reqs;
 317         ssize_t bytes;
 318         size_t size;
 319         __u64 offset;
 320         bool write;
 321         bool should_dirty;
 322         int err;
 323         struct kiocb *iocb;
 324         struct completion *done;
 325         bool blocking;
 326 };
 327
 328 #define FUSE_IO_PRIV_SYNC(i) \
 329 {                                       \
 330         .refcnt = KREF_INIT(1),         \
 331         .async = 0,                     \
 332         .iocb = i,                      \
 333 }
 334
 335 /**
 336  * Request flags
 337  *
 338  * FR_ISREPLY:          set if the request has reply
 339  * FR_FORCE:            force sending of the request even if interrupted
 340  * FR_BACKGROUND:       request is sent in the background
 341  * FR_WAITING:          request is counted as "waiting"
 342  * FR_ABORTED:          the request was aborted
 343  * FR_INTERRUPTED:      the request has been interrupted
 344  * FR_LOCKED:           data is being copied to/from the request
 345  * FR_PENDING:          request is not yet in userspace
 346  * FR_SENT:             request is in userspace, waiting for an answer
 347  * FR_FINISHED:         request is finished
 348  * FR_PRIVATE:          request is on private list
 349  * FR_ASYNC:            request is asynchronous
 350  */
 351 enum fuse_req_flag {
 352         FR_ISREPLY,
 353         FR_FORCE,
 354         FR_BACKGROUND,
 355         FR_WAITING,
 356         FR_ABORTED,
 357         FR_INTERRUPTED,
 358         FR_LOCKED,
 359         FR_PENDING,
 360         FR_SENT,
 361         FR_FINISHED,
 362         FR_PRIVATE,
 363         FR_ASYNC,
 364 };
 365
 366 /**
 367  * A request to the client
 368  *
 369  * .waitq.lock protects the following fields:
 370  *   - FR_ABORTED
 371  *   - FR_LOCKED (may also be modified under fc->lock, tested under both)
 372  */
 373 struct fuse_req {
 374         /** This can be on either pending processing or io lists in
 375             fuse_conn */
 376         struct list_head list;
 377
 378         /** Entry on the interrupts list  */
 379         struct list_head intr_entry;
 380
 381         /* Input/output arguments */
 382         struct fuse_args *args;
 383
 384         /** refcount */
 385         refcount_t count;
 386
 387         /* Request flags, updated with test/set/clear_bit() */
 388         unsigned long flags;
 389
 390         /* The request input header */
 391         struct {
 392                 struct fuse_in_header h;
 393         } in;
 394
 395         /* The request output header */
 396         struct {
 397                 struct fuse_out_header h;
 398         } out;
 399
 400         /** Used to wake up the task waiting for completion of request*/
 401         wait_queue_head_t waitq;
 402
 403 #if IS_ENABLED(CONFIG_VIRTIO_FS)
 404         /** virtio-fs's physically contiguous buffer for in and out args */
 405         void *argbuf;
 406 #endif
 407
 408         /** fuse_mount this request belongs to */
 409         struct fuse_mount *fm;
 410 };
 411
 412 struct fuse_iqueue;
 413
 414 /**
 415  * Input queue callbacks
 416  *
 417  * Input queue signalling is device-specific.  For example, the /dev/fuse file
 418  * uses fiq->waitq and fasync to wake processes that are waiting on queue
 419  * readiness.  These callbacks allow other device types to respond to input
 420  * queue activity.
 421  */
 422 struct fuse_iqueue_ops {
 423         /**
 424          * Signal that a forget has been queued
 425          */
 426         void (*wake_forget_and_unlock)(struct fuse_iqueue *fiq)
 427                 __releases(fiq->lock);
 428
 429         /**
 430          * Signal that an INTERRUPT request has been queued
 431          */
 432         void (*wake_interrupt_and_unlock)(struct fuse_iqueue *fiq)
 433                 __releases(fiq->lock);
 434
 435         /**
 436          * Signal that a request has been queued
 437          */
 438         void (*wake_pending_and_unlock)(struct fuse_iqueue *fiq)
 439                 __releases(fiq->lock);
 440
 441         /**
 442          * Clean up when fuse_iqueue is destroyed
 443          */
 444         void (*release)(struct fuse_iqueue *fiq);
 445 };
 446
 447 /** /dev/fuse input queue operations */
 448 extern const struct fuse_iqueue_ops fuse_dev_fiq_ops;
 449
 450 struct fuse_iqueue {
 451         /** Connection established */
 452         unsigned connected;
 453
 454         /** Lock protecting accesses to members of this structure */
 455         spinlock_t lock;
 456
 457         /** Readers of the connection are waiting on this */
 458         wait_queue_head_t waitq;
 459
 460         /** The next unique request id */
 461         u64 reqctr;
 462
 463         /** The list of pending requests */
 464         struct list_head pending;
 465
 466         /** Pending interrupts */
 467         struct list_head interrupts;
 468
 469         /** Queue of pending forgets */
 470         struct fuse_forget_link forget_list_head;
 471         struct fuse_forget_link *forget_list_tail;
 472
 473         /** Batching of FORGET requests (positive indicates FORGET batch) */
 474         int forget_batch;
 475
 476         /** O_ASYNC requests */
 477         struct fasync_struct *fasync;
 478
 479         /** Device-specific callbacks */
 480         const struct fuse_iqueue_ops *ops;
 481
 482         /** Device-specific state */
 483         void *priv;
 484 };
 485
 486 #define FUSE_PQ_HASH_BITS 8
 487 #define FUSE_PQ_HASH_SIZE (1 << FUSE_PQ_HASH_BITS)
 488
 489 struct fuse_pqueue {
 490         /** Connection established */
 491         unsigned connected;
 492
 493         /** Lock protecting accessess to  members of this structure */
 494         spinlock_t lock;
 495
 496         /** Hash table of requests being processed */
 497         struct list_head *processing;
 498
 499         /** The list of requests under I/O */
 500         struct list_head io;
 501 };
 502
 503 /**
 504  * Fuse device instance
 505  */
 506 struct fuse_dev {
 507         /** Fuse connection for this device */
 508         struct fuse_conn *fc;
 509
 510         /** Processing queue */
 511         struct fuse_pqueue pq;
 512
 513         /** list entry on fc->devices */
 514         struct list_head entry;
 515 };
 516
 517 enum fuse_dax_mode {
 518         FUSE_DAX_INODE_DEFAULT, /* default */
 519         FUSE_DAX_ALWAYS,        /* "-o dax=always" */
 520         FUSE_DAX_NEVER,         /* "-o dax=never" */
 521         FUSE_DAX_INODE_USER,    /* "-o dax=inode" */
 522 };
 523
 524 static inline bool fuse_is_inode_dax_mode(enum fuse_dax_mode mode)
 525 {
 526         return mode == FUSE_DAX_INODE_DEFAULT || mode == FUSE_DAX_INODE_USER;
 527 }
 528
 529 struct fuse_fs_context {
 530         int fd;
 531         struct file *file;
 532         unsigned int rootmode;
 533         kuid_t user_id;
 534         kgid_t group_id;
 535         bool is_bdev:1;
 536         bool fd_present:1;
 537         bool rootmode_present:1;
 538         bool user_id_present:1;
 539         bool group_id_present:1;
 540         bool default_permissions:1;
 541         bool allow_other:1;
 542         bool destroy:1;
 543         bool no_control:1;
 544         bool no_force_umount:1;
 545         bool legacy_opts_show:1;
 546         enum fuse_dax_mode dax_mode;
 547         unsigned int max_read;
 548         unsigned int blksize;
 549         const char *subtype;
 550
 551         /* DAX device, may be NULL */
 552         struct dax_device *dax_dev;
 553
 554         /* fuse_dev pointer to fill in, should contain NULL on entry */
 555         void **fudptr;
 556 };
 557
 558 struct fuse_sync_bucket {
 559         /* count is a possible scalability bottleneck */
 560         atomic_t count;
 561         wait_queue_head_t waitq;
 562         struct rcu_head rcu;
 563 };
 564
 565 /**
 566  * A Fuse connection.
 567  *
 568  * This structure is created, when the root filesystem is mounted, and
 569  * is destroyed, when the client device is closed and the last
 570  * fuse_mount is destroyed.
 571  */
 572 struct fuse_conn {
 573         /** Lock protecting accessess to  members of this structure */
 574         spinlock_t lock;
 575
 576         /** Refcount */
 577         refcount_t count;
 578
 579         /** Number of fuse_dev's */
 580         atomic_t dev_count;
 581
 582         struct rcu_head rcu;
 583
 584         /** The user id for this mount */
 585         kuid_t user_id;
 586
 587         /** The group id for this mount */
 588         kgid_t group_id;
 589
 590         /** The pid namespace for this mount */
 591         struct pid_namespace *pid_ns;
 592
 593         /** The user namespace for this mount */
 594         struct user_namespace *user_ns;
 595
 596         /** Maximum read size */
 597         unsigned max_read;
 598
 599         /** Maximum write size */
 600         unsigned max_write;
 601
 602         /** Maximum number of pages that can be used in a single request */
 603         unsigned int max_pages;
 604
 605         /** Constrain ->max_pages to this value during feature negotiation */
 606         unsigned int max_pages_limit;
 607
 608         /** Input queue */
 609         struct fuse_iqueue iq;
 610
 611         /** The next unique kernel file handle */
 612         atomic64_t khctr;
 613
 614         /** rbtree of fuse_files waiting for poll events indexed by ph */
 615         struct rb_root polled_files;
 616
 617         /** Maximum number of outstanding background requests */
 618         unsigned max_background;
 619
 620         /** Number of background requests at which congestion starts */
 621         unsigned congestion_threshold;
 622
 623         /** Number of requests currently in the background */
 624         unsigned num_background;
 625
 626         /** Number of background requests currently queued for userspace */
 627         unsigned active_background;
 628
 629         /** The list of background requests set aside for later queuing */
 630         struct list_head bg_queue;
 631
 632         /** Protects: max_background, congestion_threshold, num_background,
 633          * active_background, bg_queue, blocked */
 634         spinlock_t bg_lock;
 635
 636         /** Flag indicating that INIT reply has been received. Allocating
 637          * any fuse request will be suspended until the flag is set */
 638         int initialized;
 639
 640         /** Flag indicating if connection is blocked.  This will be
 641             the case before the INIT reply is received, and if there
 642             are too many outstading backgrounds requests */
 643         int blocked;
 644
 645         /** waitq for blocked connection */
 646         wait_queue_head_t blocked_waitq;
 647
 648         /** Connection established, cleared on umount, connection
 649             abort and device release */
 650         unsigned connected;
 651
 652         /** Connection aborted via sysfs */
 653         bool aborted;
 654
 655         /** Connection failed (version mismatch).  Cannot race with
 656             setting other bitfields since it is only set once in INIT
 657             reply, before any other request, and never cleared */
 658         unsigned conn_error:1;
 659
 660         /** Connection successful.  Only set in INIT */
 661         unsigned conn_init:1;
 662
 663         /** Do readahead asynchronously?  Only set in INIT */
 664         unsigned async_read:1;
 665
 666         /** Return an unique read error after abort.  Only set in INIT */
 667         unsigned abort_err:1;
 668
 669         /** Do not send separate SETATTR request before open(O_TRUNC)  */
 670         unsigned atomic_o_trunc:1;
 671
 672         /** Filesystem supports NFS exporting.  Only set in INIT */
 673         unsigned export_support:1;
 674
 675         /** write-back cache policy (default is write-through) */
 676         unsigned writeback_cache:1;
 677
 678         /** allow parallel lookups and readdir (default is serialized) */
 679         unsigned parallel_dirops:1;
 680
 681         /** handle fs handles killing suid/sgid/cap on write/chown/trunc */
 682         unsigned handle_killpriv:1;
 683
 684         /** cache READLINK responses in page cache */
 685         unsigned cache_symlinks:1;
 686
 687         /* show legacy mount options */
 688         unsigned int legacy_opts_show:1;
 689
 690         /*
 691          * fs kills suid/sgid/cap on write/chown/trunc. suid is killed on
 692          * write/trunc only if caller did not have CAP_FSETID.  sgid is killed
 693          * on write/truncate only if caller did not have CAP_FSETID as well as
 694          * file has group execute permission.
 695          */
 696         unsigned handle_killpriv_v2:1;
 697
 698         /*
 699          * The following bitfields are only for optimization purposes
 700          * and hence races in setting them will not cause malfunction
 701          */
 702
 703         /** Is open/release not implemented by fs? */
 704         unsigned no_open:1;
 705
 706         /** Is opendir/releasedir not implemented by fs? */
 707         unsigned no_opendir:1;
 708
 709         /** Is fsync not implemented by fs? */
 710         unsigned no_fsync:1;
 711
 712         /** Is fsyncdir not implemented by fs? */
 713         unsigned no_fsyncdir:1;
 714
 715         /** Is flush not implemented by fs? */
 716         unsigned no_flush:1;
 717
 718         /** Is setxattr not implemented by fs? */
 719         unsigned no_setxattr:1;
 720
 721         /** Does file server support extended setxattr */
 722         unsigned setxattr_ext:1;
 723
 724         /** Is getxattr not implemented by fs? */
 725         unsigned no_getxattr:1;
 726
 727         /** Is listxattr not implemented by fs? */
 728         unsigned no_listxattr:1;
 729
 730         /** Is removexattr not implemented by fs? */
 731         unsigned no_removexattr:1;
 732
 733         /** Are posix file locking primitives not implemented by fs? */
 734         unsigned no_lock:1;
 735
 736         /** Is access not implemented by fs? */
 737         unsigned no_access:1;
 738
 739         /** Is create not implemented by fs? */
 740         unsigned no_create:1;
 741
 742         /** Is interrupt not implemented by fs? */
 743         unsigned no_interrupt:1;
 744
 745         /** Is bmap not implemented by fs? */
 746         unsigned no_bmap:1;
 747
 748         /** Is poll not implemented by fs? */
 749         unsigned no_poll:1;
 750
 751         /** Do multi-page cached writes */
 752         unsigned big_writes:1;
 753
 754         /** Don't apply umask to creation modes */
 755         unsigned dont_mask:1;
 756
 757         /** Are BSD file locking primitives not implemented by fs? */
 758         unsigned no_flock:1;
 759
 760         /** Is fallocate not implemented by fs? */
 761         unsigned no_fallocate:1;
 762
 763         /** Is rename with flags implemented by fs? */
 764         unsigned no_rename2:1;
 765
 766         /** Use enhanced/automatic page cache invalidation. */
 767         unsigned auto_inval_data:1;
 768
 769         /** Filesystem is fully responsible for page cache invalidation. */
 770         unsigned explicit_inval_data:1;
 771
 772         /** Does the filesystem support readdirplus? */
 773         unsigned do_readdirplus:1;
 774
 775         /** Does the filesystem want adaptive readdirplus? */
 776         unsigned readdirplus_auto:1;
 777
 778         /** Does the filesystem support asynchronous direct-IO submission? */
 779         unsigned async_dio:1;
 780
 781         /** Is lseek not implemented by fs? */
 782         unsigned no_lseek:1;
 783
 784         /** Does the filesystem support posix acls? */
 785         unsigned posix_acl:1;
 786
 787         /** Check permissions based on the file mode or not? */
 788         unsigned default_permissions:1;
 789
 790         /** Allow other than the mounter user to access the filesystem ? */
 791         unsigned allow_other:1;
 792
 793         /** Does the filesystem support copy_file_range? */
 794         unsigned no_copy_file_range:1;
 795
 796         /* Send DESTROY request */
 797         unsigned int destroy:1;
 798
 799         /* Delete dentries that have gone stale */
 800         unsigned int delete_stale:1;
 801
 802         /** Do not create entry in fusectl fs */
 803         unsigned int no_control:1;
 804
 805         /** Do not allow MNT_FORCE umount */
 806         unsigned int no_force_umount:1;
 807
 808         /* Auto-mount submounts announced by the server */
 809         unsigned int auto_submounts:1;
 810
 811         /* Propagate syncfs() to server */
 812         unsigned int sync_fs:1;
 813
 814         /* Initialize security xattrs when creating a new inode */
 815         unsigned int init_security:1;
 816
 817         /* Add supplementary group info when creating a new inode */
 818         unsigned int create_supp_group:1;
 819
 820         /* Does the filesystem support per inode DAX? */
 821         unsigned int inode_dax:1;
 822
 823         /* Is tmpfile not implemented by fs? */
 824         unsigned int no_tmpfile:1;
 825
 826         /* Relax restrictions to allow shared mmap in FOPEN_DIRECT_IO mode */
 827         unsigned int direct_io_allow_mmap:1;
 828
 829         /* Is statx not implemented by fs? */
 830         unsigned int no_statx:1;
 831
 832         /** The number of requests waiting for completion */
 833         atomic_t num_waiting;
 834
 835         /** Negotiated minor version */
 836         unsigned minor;
 837
 838         /** Entry on the fuse_mount_list */
 839         struct list_head entry;
 840
 841         /** Device ID from the root super block */
 842         dev_t dev;
 843
 844         /** Dentries in the control filesystem */
 845         struct dentry *ctl_dentry[FUSE_CTL_NUM_DENTRIES];
 846
 847         /** number of dentries used in the above array */
 848         int ctl_ndents;
 849
 850         /** Key for lock owner ID scrambling */
 851         u32 scramble_key[4];
 852
 853         /** Version counter for attribute changes */
 854         atomic64_t attr_version;
 855
 856         /** Called on final put */
 857         void (*release)(struct fuse_conn *);
 858
 859         /**
 860          * Read/write semaphore to hold when accessing the sb of any
 861          * fuse_mount belonging to this connection
 862          */
 863         struct rw_semaphore killsb;
 864
 865         /** List of device instances belonging to this connection */
 866         struct list_head devices;
 867
 868 #ifdef CONFIG_FUSE_DAX
 869         /* Dax mode */
 870         enum fuse_dax_mode dax_mode;
 871
 872         /* Dax specific conn data, non-NULL if DAX is enabled */
 873         struct fuse_conn_dax *dax;
 874 #endif
 875
 876         /** List of filesystems using this connection */
 877         struct list_head mounts;
 878
 879         /* New writepages go into this bucket */
 880         struct fuse_sync_bucket __rcu *curr_bucket;
 881 };
 882
 883 /*
 884  * Represents a mounted filesystem, potentially a submount.
 885  *
 886  * This object allows sharing a fuse_conn between separate mounts to
 887  * allow submounts with dedicated superblocks and thus separate device
 888  * IDs.
 889  */
 890 struct fuse_mount {
 891         /* Underlying (potentially shared) connection to the FUSE server */
 892         struct fuse_conn *fc;
 893
 894         /*
 895          * Super block for this connection (fc->killsb must be held when
 896          * accessing this).
 897          */
 898         struct super_block *sb;
 899
 900         /* Entry on fc->mounts */
 901         struct list_head fc_entry;
 902 };
 903
 904 static inline struct fuse_mount *get_fuse_mount_super(struct super_block *sb)
 905 {
 906         return sb->s_fs_info;
 907 }
 908
 909 static inline struct fuse_conn *get_fuse_conn_super(struct super_block *sb)
 910 {
 911         return get_fuse_mount_super(sb)->fc;
 912 }
 913
 914 static inline struct fuse_mount *get_fuse_mount(struct inode *inode)
 915 {
 916         return get_fuse_mount_super(inode->i_sb);
 917 }
 918
 919 static inline struct fuse_conn *get_fuse_conn(struct inode *inode)
 920 {
 921         return get_fuse_mount_super(inode->i_sb)->fc;
 922 }
 923
 924 static inline struct fuse_inode *get_fuse_inode(struct inode *inode)
 925 {
 926         return container_of(inode, struct fuse_inode, inode);
 927 }
 928
 929 static inline u64 get_node_id(struct inode *inode)
 930 {
 931         return get_fuse_inode(inode)->nodeid;
 932 }
 933
 934 static inline int invalid_nodeid(u64 nodeid)
 935 {
 936         return !nodeid || nodeid == FUSE_ROOT_ID;
 937 }
 938
 939 static inline u64 fuse_get_attr_version(struct fuse_conn *fc)
 940 {
 941         return atomic64_read(&fc->attr_version);
 942 }
 943
 944 static inline bool fuse_stale_inode(const struct inode *inode, int generation,
 945                                     struct fuse_attr *attr)
 946 {
 947         return inode->i_generation != generation ||
 948                 inode_wrong_type(inode, attr->mode);
 949 }
 950
 951 static inline void fuse_make_bad(struct inode *inode)
 952 {
 953         remove_inode_hash(inode);
 954         set_bit(FUSE_I_BAD, &get_fuse_inode(inode)->state);
 955 }
 956
 957 static inline bool fuse_is_bad(struct inode *inode)
 958 {
 959         return unlikely(test_bit(FUSE_I_BAD, &get_fuse_inode(inode)->state));
 960 }
 961
 962 static inline struct page **fuse_pages_alloc(unsigned int npages, gfp_t flags,
 963                                              struct fuse_page_desc **desc)
 964 {
 965         struct page **pages;
 966
 967         pages = kzalloc(npages * (sizeof(struct page *) +
 968                                   sizeof(struct fuse_page_desc)), flags);
 969         *desc = (void *) (pages + npages);
 970
 971         return pages;
 972 }
 973
 974 static inline void fuse_page_descs_length_init(struct fuse_page_desc *descs,
 975                                                unsigned int index,
 976                                                unsigned int nr_pages)
 977 {
 978         int i;
 979
 980         for (i = index; i < index + nr_pages; i++)
 981                 descs[i].length = PAGE_SIZE - descs[i].offset;
 982 }
 983
 984 static inline void fuse_sync_bucket_dec(struct fuse_sync_bucket *bucket)
 985 {
 986         /* Need RCU protection to prevent use after free after the decrement */
 987         rcu_read_lock();
 988         if (atomic_dec_and_test(&bucket->count))
 989                 wake_up(&bucket->waitq);
 990         rcu_read_unlock();
 991 }
 992
 993 /** Device operations */
 994 extern const struct file_operations fuse_dev_operations;
 995
 996 extern const struct dentry_operations fuse_dentry_operations;
 997 extern const struct dentry_operations fuse_root_dentry_operations;
 998
 999 /**
1000  * Get a filled in inode
1001  */
1002 struct inode *fuse_iget(struct super_block *sb, u64 nodeid,
1003                         int generation, struct fuse_attr *attr,
1004                         u64 attr_valid, u64 attr_version);
1005
1006 int fuse_lookup_name(struct super_block *sb, u64 nodeid, const struct qstr *name,
1007                      struct fuse_entry_out *outarg, struct inode **inode);
1008
1009 /**
1010  * Send FORGET command
1011  */
1012 void fuse_queue_forget(struct fuse_conn *fc, struct fuse_forget_link *forget,
1013                        u64 nodeid, u64 nlookup);
1014
1015 struct fuse_forget_link *fuse_alloc_forget(void);
1016
1017 struct fuse_forget_link *fuse_dequeue_forget(struct fuse_iqueue *fiq,
1018                                              unsigned int max,
1019                                              unsigned int *countp);
1020
1021 /*
1022  * Initialize READ or READDIR request
1023  */
1024 struct fuse_io_args {
1025         union {
1026                 struct {
1027                         struct fuse_read_in in;
1028                         u64 attr_ver;
1029                 } read;
1030                 struct {
1031                         struct fuse_write_in in;
1032                         struct fuse_write_out out;
1033                         bool page_locked;
1034                 } write;
1035         };
1036         struct fuse_args_pages ap;
1037         struct fuse_io_priv *io;
1038         struct fuse_file *ff;
1039 };
1040
1041 void fuse_read_args_fill(struct fuse_io_args *ia, struct file *file, loff_t pos,
1042                          size_t count, int opcode);
1043
1044
1045 struct fuse_file *fuse_file_alloc(struct fuse_mount *fm, bool release);
1046 void fuse_file_free(struct fuse_file *ff);
1047 int fuse_finish_open(struct inode *inode, struct file *file);
1048
1049 void fuse_sync_release(struct fuse_inode *fi, struct fuse_file *ff,
1050                        unsigned int flags);
1051
1052 /**
1053  * Send RELEASE or RELEASEDIR request
1054  */
1055 void fuse_release_common(struct file *file, bool isdir);
1056
1057 /**
1058  * Send FSYNC or FSYNCDIR request
1059  */
1060 int fuse_fsync_common(struct file *file, loff_t start, loff_t end,
1061                       int datasync, int opcode);
1062
1063 /**
1064  * Notify poll wakeup
1065  */
1066 int fuse_notify_poll_wakeup(struct fuse_conn *fc,
1067                             struct fuse_notify_poll_wakeup_out *outarg);
1068
1069 /**
1070  * Initialize file operations on a regular file
1071  */
1072 void fuse_init_file_inode(struct inode *inode, unsigned int flags);
1073
1074 /**
1075  * Initialize inode operations on regular files and special files
1076  */
1077 void fuse_init_common(struct inode *inode);
1078
1079 /**
1080  * Initialize inode and file operations on a directory
1081  */
1082 void fuse_init_dir(struct inode *inode);
1083
1084 /**
1085  * Initialize inode operations on a symlink
1086  */
1087 void fuse_init_symlink(struct inode *inode);
1088
1089 /**
1090  * Change attributes of an inode
1091  */
1092 void fuse_change_attributes(struct inode *inode, struct fuse_attr *attr,
1093                             struct fuse_statx *sx,
1094                             u64 attr_valid, u64 attr_version);
1095
1096 void fuse_change_attributes_common(struct inode *inode, struct fuse_attr *attr,
1097                                    struct fuse_statx *sx,
1098                                    u64 attr_valid, u32 cache_mask);
1099
1100 u32 fuse_get_cache_mask(struct inode *inode);
1101
1102 /**
1103  * Initialize the client device
1104  */
1105 int fuse_dev_init(void);
1106
1107 /**
1108  * Cleanup the client device
1109  */
1110 void fuse_dev_cleanup(void);
1111
1112 int fuse_ctl_init(void);
1113 void __exit fuse_ctl_cleanup(void);
1114
1115 /**
1116  * Simple request sending that does request allocation and freeing
1117  */
1118 ssize_t fuse_simple_request(struct fuse_mount *fm, struct fuse_args *args);
1119 int fuse_simple_background(struct fuse_mount *fm, struct fuse_args *args,
1120                            gfp_t gfp_flags);
1121
1122 /**
1123  * End a finished request
1124  */
1125 void fuse_request_end(struct fuse_req *req);
1126
1127 /* Abort all requests */
1128 void fuse_abort_conn(struct fuse_conn *fc);
1129 void fuse_wait_aborted(struct fuse_conn *fc);
1130
1131 /**
1132  * Invalidate inode attributes
1133  */
1134
1135 /* Attributes possibly changed on data modification */
1136 #define FUSE_STATX_MODIFY       (STATX_MTIME | STATX_CTIME | STATX_BLOCKS)
1137
1138 /* Attributes possibly changed on data and/or size modification */
1139 #define FUSE_STATX_MODSIZE      (FUSE_STATX_MODIFY | STATX_SIZE)
1140
1141 void fuse_invalidate_attr(struct inode *inode);
1142 void fuse_invalidate_attr_mask(struct inode *inode, u32 mask);
1143
1144 void fuse_invalidate_entry_cache(struct dentry *entry);
1145
1146 void fuse_invalidate_atime(struct inode *inode);
1147
1148 u64 fuse_time_to_jiffies(u64 sec, u32 nsec);
1149 #define ATTR_TIMEOUT(o) \
1150         fuse_time_to_jiffies((o)->attr_valid, (o)->attr_valid_nsec)
1151
1152 void fuse_change_entry_timeout(struct dentry *entry, struct fuse_entry_out *o);
1153
1154 /**
1155  * Acquire reference to fuse_conn
1156  */
1157 struct fuse_conn *fuse_conn_get(struct fuse_conn *fc);
1158
1159 /**
1160  * Initialize fuse_conn
1161  */
1162 void fuse_conn_init(struct fuse_conn *fc, struct fuse_mount *fm,
1163                     struct user_namespace *user_ns,
1164                     const struct fuse_iqueue_ops *fiq_ops, void *fiq_priv);
1165
1166 /**
1167  * Release reference to fuse_conn
1168  */
1169 void fuse_conn_put(struct fuse_conn *fc);
1170
1171 struct fuse_dev *fuse_dev_alloc_install(struct fuse_conn *fc);
1172 struct fuse_dev *fuse_dev_alloc(void);
1173 void fuse_dev_install(struct fuse_dev *fud, struct fuse_conn *fc);
1174 void fuse_dev_free(struct fuse_dev *fud);
1175 void fuse_send_init(struct fuse_mount *fm);
1176
1177 /**
1178  * Fill in superblock and initialize fuse connection
1179  * @sb: partially-initialized superblock to fill in
1180  * @ctx: mount context
1181  */
1182 int fuse_fill_super_common(struct super_block *sb, struct fuse_fs_context *ctx);
1183
1184 /*
1185  * Remove the mount from the connection
1186  *
1187  * Returns whether this was the last mount
1188  */
1189 bool fuse_mount_remove(struct fuse_mount *fm);
1190
1191 /*
1192  * Setup context ops for submounts
1193  */
1194 int fuse_init_fs_context_submount(struct fs_context *fsc);
1195
1196 /*
1197  * Shut down the connection (possibly sending DESTROY request).
1198  */
1199 void fuse_conn_destroy(struct fuse_mount *fm);
1200
1201 /* Drop the connection and free the fuse mount */
1202 void fuse_mount_destroy(struct fuse_mount *fm);
1203
1204 /**
1205  * Add connection to control filesystem
1206  */
1207 int fuse_ctl_add_conn(struct fuse_conn *fc);
1208
1209 /**
1210  * Remove connection from control filesystem
1211  */
1212 void fuse_ctl_remove_conn(struct fuse_conn *fc);
1213
1214 /**
1215  * Is file type valid?
1216  */
1217 int fuse_valid_type(int m);
1218
1219 bool fuse_invalid_attr(struct fuse_attr *attr);
1220
1221 /**
1222  * Is current process allowed to perform filesystem operation?
1223  */
1224 bool fuse_allow_current_process(struct fuse_conn *fc);
1225
1226 u64 fuse_lock_owner_id(struct fuse_conn *fc, fl_owner_t id);
1227
1228 void fuse_flush_time_update(struct inode *inode);
1229 void fuse_update_ctime(struct inode *inode);
1230
1231 int fuse_update_attributes(struct inode *inode, struct file *file, u32 mask);
1232
1233 void fuse_flush_writepages(struct inode *inode);
1234
1235 void fuse_set_nowrite(struct inode *inode);
1236 void fuse_release_nowrite(struct inode *inode);
1237
1238 /**
1239  * Scan all fuse_mounts belonging to fc to find the first where
1240  * ilookup5() returns a result.  Return that result and the
1241  * respective fuse_mount in *fm (unless fm is NULL).
1242  *
1243  * The caller must hold fc->killsb.
1244  */
1245 struct inode *fuse_ilookup(struct fuse_conn *fc, u64 nodeid,
1246                            struct fuse_mount **fm);
1247
1248 /**
1249  * File-system tells the kernel to invalidate cache for the given node id.
1250  */
1251 int fuse_reverse_inval_inode(struct fuse_conn *fc, u64 nodeid,
1252                              loff_t offset, loff_t len);
1253
1254 /**
1255  * File-system tells the kernel to invalidate parent attributes and
1256  * the dentry matching parent/name.
1257  *
1258  * If the child_nodeid is non-zero and:
1259  *    - matches the inode number for the dentry matching parent/name,
1260  *    - is not a mount point
1261  *    - is a file or oan empty directory
1262  * then the dentry is unhashed (d_delete()).
1263  */
1264 int fuse_reverse_inval_entry(struct fuse_conn *fc, u64 parent_nodeid,
1265                              u64 child_nodeid, struct qstr *name, u32 flags);
1266
1267 int fuse_do_open(struct fuse_mount *fm, u64 nodeid, struct file *file,
1268                  bool isdir);
1269
1270 /**
1271  * fuse_direct_io() flags
1272  */
1273
1274 /** If set, it is WRITE; otherwise - READ */
1275 #define FUSE_DIO_WRITE (1 << 0)
1276
1277 /** CUSE pass fuse_direct_io() a file which f_mapping->host is not from FUSE */
1278 #define FUSE_DIO_CUSE  (1 << 1)
1279
1280 ssize_t fuse_direct_io(struct fuse_io_priv *io, struct iov_iter *iter,
1281                        loff_t *ppos, int flags);
1282 long fuse_do_ioctl(struct file *file, unsigned int cmd, unsigned long arg,
1283                    unsigned int flags);
1284 long fuse_ioctl_common(struct file *file, unsigned int cmd,
1285                        unsigned long arg, unsigned int flags);
1286 __poll_t fuse_file_poll(struct file *file, poll_table *wait);
1287 int fuse_dev_release(struct inode *inode, struct file *file);
1288
1289 bool fuse_write_update_attr(struct inode *inode, loff_t pos, ssize_t written);
1290
1291 int fuse_flush_times(struct inode *inode, struct fuse_file *ff);
1292 int fuse_write_inode(struct inode *inode, struct writeback_control *wbc);
1293
1294 int fuse_do_setattr(struct dentry *dentry, struct iattr *attr,
1295                     struct file *file);
1296
1297 void fuse_set_initialized(struct fuse_conn *fc);
1298
1299 void fuse_unlock_inode(struct inode *inode, bool locked);
1300 bool fuse_lock_inode(struct inode *inode);
1301
1302 int fuse_setxattr(struct inode *inode, const char *name, const void *value,
1303                   size_t size, int flags, unsigned int extra_flags);
1304 ssize_t fuse_getxattr(struct inode *inode, const char *name, void *value,
1305                       size_t size);
1306 ssize_t fuse_listxattr(struct dentry *entry, char *list, size_t size);
1307 int fuse_removexattr(struct inode *inode, const char *name);
1308 extern const struct xattr_handler * const fuse_xattr_handlers[];
1309
1310 struct posix_acl;
1311 struct posix_acl *fuse_get_inode_acl(struct inode *inode, int type, bool rcu);
1312 struct posix_acl *fuse_get_acl(struct mnt_idmap *idmap,
1313                                struct dentry *dentry, int type);
1314 int fuse_set_acl(struct mnt_idmap *, struct dentry *dentry,
1315                  struct posix_acl *acl, int type);
1316
1317 /* readdir.c */
1318 int fuse_readdir(struct file *file, struct dir_context *ctx);
1319
1320 /**
1321  * Return the number of bytes in an arguments list
1322  */
1323 unsigned int fuse_len_args(unsigned int numargs, struct fuse_arg *args);
1324
1325 /**
1326  * Get the next unique ID for a request
1327  */
1328 u64 fuse_get_unique(struct fuse_iqueue *fiq);
1329 void fuse_free_conn(struct fuse_conn *fc);
1330
1331 /* dax.c */
1332
1333 #define FUSE_IS_DAX(inode) (IS_ENABLED(CONFIG_FUSE_DAX) && IS_DAX(inode))
1334
1335 ssize_t fuse_dax_read_iter(struct kiocb *iocb, struct iov_iter *to);
1336 ssize_t fuse_dax_write_iter(struct kiocb *iocb, struct iov_iter *from);
1337 int fuse_dax_mmap(struct file *file, struct vm_area_struct *vma);
1338 int fuse_dax_break_layouts(struct inode *inode, u64 dmap_start, u64 dmap_end);
1339 int fuse_dax_conn_alloc(struct fuse_conn *fc, enum fuse_dax_mode mode,
1340                         struct dax_device *dax_dev);
1341 void fuse_dax_conn_free(struct fuse_conn *fc);
1342 bool fuse_dax_inode_alloc(struct super_block *sb, struct fuse_inode *fi);
1343 void fuse_dax_inode_init(struct inode *inode, unsigned int flags);
1344 void fuse_dax_inode_cleanup(struct inode *inode);
1345 void fuse_dax_dontcache(struct inode *inode, unsigned int flags);
1346 bool fuse_dax_check_alignment(struct fuse_conn *fc, unsigned int map_alignment);
1347 void fuse_dax_cancel_work(struct fuse_conn *fc);
1348
1349 /* ioctl.c */
1350 long fuse_file_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
1351 long fuse_file_compat_ioctl(struct file *file, unsigned int cmd,
1352                             unsigned long arg);
1353 int fuse_fileattr_get(struct dentry *dentry, struct fileattr *fa);
1354 int fuse_fileattr_set(struct mnt_idmap *idmap,
1355                       struct dentry *dentry, struct fileattr *fa);
1356
1357 /* iomode.c */
1358 int fuse_file_cached_io_start(struct inode *inode, struct fuse_file *ff);
1359 int fuse_file_uncached_io_start(struct inode *inode, struct fuse_file *ff);
1360 void fuse_file_uncached_io_end(struct inode *inode, struct fuse_file *ff);
1361
1362 int fuse_file_io_open(struct file *file, struct inode *inode);
1363 void fuse_file_io_release(struct fuse_file *ff, struct inode *inode);
1364
1365 /* file.c */
1366 struct fuse_file *fuse_file_open(struct fuse_mount *fm, u64 nodeid,
1367                                  unsigned int open_flags, bool isdir);
1368 void fuse_file_release(struct inode *inode, struct fuse_file *ff,
1369                        unsigned int open_flags, fl_owner_t id, bool isdir);
1370
1371 #endif /* _FS_FUSE_I_H */