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 /** Container for data related to mapping to backing file */
  80 struct fuse_backing {
  81         struct file *file;
  82
  83         /** refcount */
  84         refcount_t count;
  85         struct rcu_head rcu;
  86 };
  87
  88 /** FUSE inode */
  89 struct fuse_inode {
  90         /** Inode data */
  91         struct inode inode;
  92
  93         /** Unique ID, which identifies the inode between userspace
  94          * and kernel */
  95         u64 nodeid;
  96
  97         /** Number of lookups on this inode */
  98         u64 nlookup;
  99
 100         /** The request used for sending the FORGET message */
 101         struct fuse_forget_link *forget;
 102
 103         /** Time in jiffies until the file attributes are valid */
 104         u64 i_time;
 105
 106         /* Which attributes are invalid */
 107         u32 inval_mask;
 108
 109         /** The sticky bit in inode->i_mode may have been removed, so
 110             preserve the original mode */
 111         umode_t orig_i_mode;
 112
 113         /* Cache birthtime */
 114         struct timespec64 i_btime;
 115
 116         /** 64 bit inode number */
 117         u64 orig_ino;
 118
 119         /** Version of last attribute change */
 120         u64 attr_version;
 121
 122         union {
 123                 /* read/write io cache (regular file only) */
 124                 struct {
 125                         /* Files usable in writepage.  Protected by fi->lock */
 126                         struct list_head write_files;
 127
 128                         /* Writepages pending on truncate or fsync */
 129                         struct list_head queued_writes;
 130
 131                         /* Number of sent writes, a negative bias
 132                          * (FUSE_NOWRITE) means more writes are blocked */
 133                         int writectr;
 134
 135                         /** Number of files/maps using page cache */
 136                         int iocachectr;
 137
 138                         /* Waitq for writepage completion */
 139                         wait_queue_head_t page_waitq;
 140
 141                         /* waitq for direct-io completion */
 142                         wait_queue_head_t direct_io_waitq;
 143
 144                         /* List of writepage requestst (pending or sent) */
 145                         struct rb_root writepages;
 146                 };
 147
 148                 /* readdir cache (directory only) */
 149                 struct {
 150                         /* true if fully cached */
 151                         bool cached;
 152
 153                         /* size of cache */
 154                         loff_t size;
 155
 156                         /* position at end of cache (position of next entry) */
 157                         loff_t pos;
 158
 159                         /* version of the cache */
 160                         u64 version;
 161
 162                         /* modification time of directory when cache was
 163                          * started */
 164                         struct timespec64 mtime;
 165
 166                         /* iversion of directory when cache was started */
 167                         u64 iversion;
 168
 169                         /* protects above fields */
 170                         spinlock_t lock;
 171                 } rdc;
 172         };
 173
 174         /** Miscellaneous bits describing inode state */
 175         unsigned long state;
 176
 177         /** Lock for serializing lookup and readdir for back compatibility*/
 178         struct mutex mutex;
 179
 180         /** Lock to protect write related fields */
 181         spinlock_t lock;
 182
 183 #ifdef CONFIG_FUSE_DAX
 184         /*
 185          * Dax specific inode data
 186          */
 187         struct fuse_inode_dax *dax;
 188 #endif
 189         /** Submount specific lookup tracking */
 190         struct fuse_submount_lookup *submount_lookup;
 191 #ifdef CONFIG_FUSE_PASSTHROUGH
 192         /** Reference to backing file in passthrough mode */
 193         struct fuse_backing *fb;
 194 #endif
 195 };
 196
 197 /** FUSE inode state bits */
 198 enum {
 199         /** Advise readdirplus  */
 200         FUSE_I_ADVISE_RDPLUS,
 201         /** Initialized with readdirplus */
 202         FUSE_I_INIT_RDPLUS,
 203         /** An operation changing file size is in progress  */
 204         FUSE_I_SIZE_UNSTABLE,
 205         /* Bad inode */
 206         FUSE_I_BAD,
 207         /* Has btime */
 208         FUSE_I_BTIME,
 209         /* Wants or already has page cache IO */
 210         FUSE_I_CACHE_IO_MODE,
 211 };
 212
 213 struct fuse_conn;
 214 struct fuse_mount;
 215 struct fuse_release_args;
 216
 217 /** FUSE specific file data */
 218 struct fuse_file {
 219         /** Fuse connection for this file */
 220         struct fuse_mount *fm;
 221
 222         /* Argument space reserved for release */
 223         struct fuse_release_args *release_args;
 224
 225         /** Kernel file handle guaranteed to be unique */
 226         u64 kh;
 227
 228         /** File handle used by userspace */
 229         u64 fh;
 230
 231         /** Node id of this file */
 232         u64 nodeid;
 233
 234         /** Refcount */
 235         refcount_t count;
 236
 237         /** FOPEN_* flags returned by open */
 238         u32 open_flags;
 239
 240         /** Entry on inode's write_files list */
 241         struct list_head write_entry;
 242
 243         /* Readdir related */
 244         struct {
 245                 /*
 246                  * Protects below fields against (crazy) parallel readdir on
 247                  * same open file.  Uncontended in the normal case.
 248                  */
 249                 struct mutex lock;
 250
 251                 /* Dir stream position */
 252                 loff_t pos;
 253
 254                 /* Offset in cache */
 255                 loff_t cache_off;
 256
 257                 /* Version of cache we are reading */
 258                 u64 version;
 259
 260         } readdir;
 261
 262         /** RB node to be linked on fuse_conn->polled_files */
 263         struct rb_node polled_node;
 264
 265         /** Wait queue head for poll */
 266         wait_queue_head_t poll_wait;
 267
 268         /** Does file hold a fi->iocachectr refcount? */
 269         enum { IOM_NONE, IOM_CACHED, IOM_UNCACHED } iomode;
 270
 271         /** Has flock been performed on this file? */
 272         bool flock:1;
 273 };
 274
 275 /** One input argument of a request */
 276 struct fuse_in_arg {
 277         unsigned size;
 278         const void *value;
 279 };
 280
 281 /** One output argument of a request */
 282 struct fuse_arg {
 283         unsigned size;
 284         void *value;
 285 };
 286
 287 /** FUSE page descriptor */
 288 struct fuse_page_desc {
 289         unsigned int length;
 290         unsigned int offset;
 291 };
 292
 293 struct fuse_args {
 294         uint64_t nodeid;
 295         uint32_t opcode;
 296         uint8_t in_numargs;
 297         uint8_t out_numargs;
 298         uint8_t ext_idx;
 299         bool force:1;
 300         bool noreply:1;
 301         bool nocreds:1;
 302         bool in_pages:1;
 303         bool out_pages:1;
 304         bool user_pages:1;
 305         bool out_argvar:1;
 306         bool page_zeroing:1;
 307         bool page_replace:1;
 308         bool may_block:1;
 309         bool is_ext:1;
 310         struct fuse_in_arg in_args[3];
 311         struct fuse_arg out_args[2];
 312         void (*end)(struct fuse_mount *fm, struct fuse_args *args, int error);
 313 };
 314
 315 struct fuse_args_pages {
 316         struct fuse_args args;
 317         struct page **pages;
 318         struct fuse_page_desc *descs;
 319         unsigned int num_pages;
 320 };
 321
 322 #define FUSE_ARGS(args) struct fuse_args args = {}
 323
 324 /** The request IO state (for asynchronous processing) */
 325 struct fuse_io_priv {
 326         struct kref refcnt;
 327         int async;
 328         spinlock_t lock;
 329         unsigned reqs;
 330         ssize_t bytes;
 331         size_t size;
 332         __u64 offset;
 333         bool write;
 334         bool should_dirty;
 335         int err;
 336         struct kiocb *iocb;
 337         struct completion *done;
 338         bool blocking;
 339 };
 340
 341 #define FUSE_IO_PRIV_SYNC(i) \
 342 {                                       \
 343         .refcnt = KREF_INIT(1),         \
 344         .async = 0,                     \
 345         .iocb = i,                      \
 346 }
 347
 348 /**
 349  * Request flags
 350  *
 351  * FR_ISREPLY:          set if the request has reply
 352  * FR_FORCE:            force sending of the request even if interrupted
 353  * FR_BACKGROUND:       request is sent in the background
 354  * FR_WAITING:          request is counted as "waiting"
 355  * FR_ABORTED:          the request was aborted
 356  * FR_INTERRUPTED:      the request has been interrupted
 357  * FR_LOCKED:           data is being copied to/from the request
 358  * FR_PENDING:          request is not yet in userspace
 359  * FR_SENT:             request is in userspace, waiting for an answer
 360  * FR_FINISHED:         request is finished
 361  * FR_PRIVATE:          request is on private list
 362  * FR_ASYNC:            request is asynchronous
 363  */
 364 enum fuse_req_flag {
 365         FR_ISREPLY,
 366         FR_FORCE,
 367         FR_BACKGROUND,
 368         FR_WAITING,
 369         FR_ABORTED,
 370         FR_INTERRUPTED,
 371         FR_LOCKED,
 372         FR_PENDING,
 373         FR_SENT,
 374         FR_FINISHED,
 375         FR_PRIVATE,
 376         FR_ASYNC,
 377 };
 378
 379 /**
 380  * A request to the client
 381  *
 382  * .waitq.lock protects the following fields:
 383  *   - FR_ABORTED
 384  *   - FR_LOCKED (may also be modified under fc->lock, tested under both)
 385  */
 386 struct fuse_req {
 387         /** This can be on either pending processing or io lists in
 388             fuse_conn */
 389         struct list_head list;
 390
 391         /** Entry on the interrupts list  */
 392         struct list_head intr_entry;
 393
 394         /* Input/output arguments */
 395         struct fuse_args *args;
 396
 397         /** refcount */
 398         refcount_t count;
 399
 400         /* Request flags, updated with test/set/clear_bit() */
 401         unsigned long flags;
 402
 403         /* The request input header */
 404         struct {
 405                 struct fuse_in_header h;
 406         } in;
 407
 408         /* The request output header */
 409         struct {
 410                 struct fuse_out_header h;
 411         } out;
 412
 413         /** Used to wake up the task waiting for completion of request*/
 414         wait_queue_head_t waitq;
 415
 416 #if IS_ENABLED(CONFIG_VIRTIO_FS)
 417         /** virtio-fs's physically contiguous buffer for in and out args */
 418         void *argbuf;
 419 #endif
 420
 421         /** fuse_mount this request belongs to */
 422         struct fuse_mount *fm;
 423 };
 424
 425 struct fuse_iqueue;
 426
 427 /**
 428  * Input queue callbacks
 429  *
 430  * Input queue signalling is device-specific.  For example, the /dev/fuse file
 431  * uses fiq->waitq and fasync to wake processes that are waiting on queue
 432  * readiness.  These callbacks allow other device types to respond to input
 433  * queue activity.
 434  */
 435 struct fuse_iqueue_ops {
 436         /**
 437          * Signal that a forget has been queued
 438          */
 439         void (*wake_forget_and_unlock)(struct fuse_iqueue *fiq)
 440                 __releases(fiq->lock);
 441
 442         /**
 443          * Signal that an INTERRUPT request has been queued
 444          */
 445         void (*wake_interrupt_and_unlock)(struct fuse_iqueue *fiq)
 446                 __releases(fiq->lock);
 447
 448         /**
 449          * Signal that a request has been queued
 450          */
 451         void (*wake_pending_and_unlock)(struct fuse_iqueue *fiq)
 452                 __releases(fiq->lock);
 453
 454         /**
 455          * Clean up when fuse_iqueue is destroyed
 456          */
 457         void (*release)(struct fuse_iqueue *fiq);
 458 };
 459
 460 /** /dev/fuse input queue operations */
 461 extern const struct fuse_iqueue_ops fuse_dev_fiq_ops;
 462
 463 struct fuse_iqueue {
 464         /** Connection established */
 465         unsigned connected;
 466
 467         /** Lock protecting accesses to members of this structure */
 468         spinlock_t lock;
 469
 470         /** Readers of the connection are waiting on this */
 471         wait_queue_head_t waitq;
 472
 473         /** The next unique request id */
 474         u64 reqctr;
 475
 476         /** The list of pending requests */
 477         struct list_head pending;
 478
 479         /** Pending interrupts */
 480         struct list_head interrupts;
 481
 482         /** Queue of pending forgets */
 483         struct fuse_forget_link forget_list_head;
 484         struct fuse_forget_link *forget_list_tail;
 485
 486         /** Batching of FORGET requests (positive indicates FORGET batch) */
 487         int forget_batch;
 488
 489         /** O_ASYNC requests */
 490         struct fasync_struct *fasync;
 491
 492         /** Device-specific callbacks */
 493         const struct fuse_iqueue_ops *ops;
 494
 495         /** Device-specific state */
 496         void *priv;
 497 };
 498
 499 #define FUSE_PQ_HASH_BITS 8
 500 #define FUSE_PQ_HASH_SIZE (1 << FUSE_PQ_HASH_BITS)
 501
 502 struct fuse_pqueue {
 503         /** Connection established */
 504         unsigned connected;
 505
 506         /** Lock protecting accessess to  members of this structure */
 507         spinlock_t lock;
 508
 509         /** Hash table of requests being processed */
 510         struct list_head *processing;
 511
 512         /** The list of requests under I/O */
 513         struct list_head io;
 514 };
 515
 516 /**
 517  * Fuse device instance
 518  */
 519 struct fuse_dev {
 520         /** Fuse connection for this device */
 521         struct fuse_conn *fc;
 522
 523         /** Processing queue */
 524         struct fuse_pqueue pq;
 525
 526         /** list entry on fc->devices */
 527         struct list_head entry;
 528 };
 529
 530 enum fuse_dax_mode {
 531         FUSE_DAX_INODE_DEFAULT, /* default */
 532         FUSE_DAX_ALWAYS,        /* "-o dax=always" */
 533         FUSE_DAX_NEVER,         /* "-o dax=never" */
 534         FUSE_DAX_INODE_USER,    /* "-o dax=inode" */
 535 };
 536
 537 static inline bool fuse_is_inode_dax_mode(enum fuse_dax_mode mode)
 538 {
 539         return mode == FUSE_DAX_INODE_DEFAULT || mode == FUSE_DAX_INODE_USER;
 540 }
 541
 542 struct fuse_fs_context {
 543         int fd;
 544         struct file *file;
 545         unsigned int rootmode;
 546         kuid_t user_id;
 547         kgid_t group_id;
 548         bool is_bdev:1;
 549         bool fd_present:1;
 550         bool rootmode_present:1;
 551         bool user_id_present:1;
 552         bool group_id_present:1;
 553         bool default_permissions:1;
 554         bool allow_other:1;
 555         bool destroy:1;
 556         bool no_control:1;
 557         bool no_force_umount:1;
 558         bool legacy_opts_show:1;
 559         enum fuse_dax_mode dax_mode;
 560         unsigned int max_read;
 561         unsigned int blksize;
 562         const char *subtype;
 563
 564         /* DAX device, may be NULL */
 565         struct dax_device *dax_dev;
 566
 567         /* fuse_dev pointer to fill in, should contain NULL on entry */
 568         void **fudptr;
 569 };
 570
 571 struct fuse_sync_bucket {
 572         /* count is a possible scalability bottleneck */
 573         atomic_t count;
 574         wait_queue_head_t waitq;
 575         struct rcu_head rcu;
 576 };
 577
 578 /**
 579  * A Fuse connection.
 580  *
 581  * This structure is created, when the root filesystem is mounted, and
 582  * is destroyed, when the client device is closed and the last
 583  * fuse_mount is destroyed.
 584  */
 585 struct fuse_conn {
 586         /** Lock protecting accessess to  members of this structure */
 587         spinlock_t lock;
 588
 589         /** Refcount */
 590         refcount_t count;
 591
 592         /** Number of fuse_dev's */
 593         atomic_t dev_count;
 594
 595         struct rcu_head rcu;
 596
 597         /** The user id for this mount */
 598         kuid_t user_id;
 599
 600         /** The group id for this mount */
 601         kgid_t group_id;
 602
 603         /** The pid namespace for this mount */
 604         struct pid_namespace *pid_ns;
 605
 606         /** The user namespace for this mount */
 607         struct user_namespace *user_ns;
 608
 609         /** Maximum read size */
 610         unsigned max_read;
 611
 612         /** Maximum write size */
 613         unsigned max_write;
 614
 615         /** Maximum number of pages that can be used in a single request */
 616         unsigned int max_pages;
 617
 618         /** Constrain ->max_pages to this value during feature negotiation */
 619         unsigned int max_pages_limit;
 620
 621         /** Input queue */
 622         struct fuse_iqueue iq;
 623
 624         /** The next unique kernel file handle */
 625         atomic64_t khctr;
 626
 627         /** rbtree of fuse_files waiting for poll events indexed by ph */
 628         struct rb_root polled_files;
 629
 630         /** Maximum number of outstanding background requests */
 631         unsigned max_background;
 632
 633         /** Number of background requests at which congestion starts */
 634         unsigned congestion_threshold;
 635
 636         /** Number of requests currently in the background */
 637         unsigned num_background;
 638
 639         /** Number of background requests currently queued for userspace */
 640         unsigned active_background;
 641
 642         /** The list of background requests set aside for later queuing */
 643         struct list_head bg_queue;
 644
 645         /** Protects: max_background, congestion_threshold, num_background,
 646          * active_background, bg_queue, blocked */
 647         spinlock_t bg_lock;
 648
 649         /** Flag indicating that INIT reply has been received. Allocating
 650          * any fuse request will be suspended until the flag is set */
 651         int initialized;
 652
 653         /** Flag indicating if connection is blocked.  This will be
 654             the case before the INIT reply is received, and if there
 655             are too many outstading backgrounds requests */
 656         int blocked;
 657
 658         /** waitq for blocked connection */
 659         wait_queue_head_t blocked_waitq;
 660
 661         /** Connection established, cleared on umount, connection
 662             abort and device release */
 663         unsigned connected;
 664
 665         /** Connection aborted via sysfs */
 666         bool aborted;
 667
 668         /** Connection failed (version mismatch).  Cannot race with
 669             setting other bitfields since it is only set once in INIT
 670             reply, before any other request, and never cleared */
 671         unsigned conn_error:1;
 672
 673         /** Connection successful.  Only set in INIT */
 674         unsigned conn_init:1;
 675
 676         /** Do readahead asynchronously?  Only set in INIT */
 677         unsigned async_read:1;
 678
 679         /** Return an unique read error after abort.  Only set in INIT */
 680         unsigned abort_err:1;
 681
 682         /** Do not send separate SETATTR request before open(O_TRUNC)  */
 683         unsigned atomic_o_trunc:1;
 684
 685         /** Filesystem supports NFS exporting.  Only set in INIT */
 686         unsigned export_support:1;
 687
 688         /** write-back cache policy (default is write-through) */
 689         unsigned writeback_cache:1;
 690
 691         /** allow parallel lookups and readdir (default is serialized) */
 692         unsigned parallel_dirops:1;
 693
 694         /** handle fs handles killing suid/sgid/cap on write/chown/trunc */
 695         unsigned handle_killpriv:1;
 696
 697         /** cache READLINK responses in page cache */
 698         unsigned cache_symlinks:1;
 699
 700         /* show legacy mount options */
 701         unsigned int legacy_opts_show:1;
 702
 703         /*
 704          * fs kills suid/sgid/cap on write/chown/trunc. suid is killed on
 705          * write/trunc only if caller did not have CAP_FSETID.  sgid is killed
 706          * on write/truncate only if caller did not have CAP_FSETID as well as
 707          * file has group execute permission.
 708          */
 709         unsigned handle_killpriv_v2:1;
 710
 711         /*
 712          * The following bitfields are only for optimization purposes
 713          * and hence races in setting them will not cause malfunction
 714          */
 715
 716         /** Is open/release not implemented by fs? */
 717         unsigned no_open:1;
 718
 719         /** Is opendir/releasedir not implemented by fs? */
 720         unsigned no_opendir:1;
 721
 722         /** Is fsync not implemented by fs? */
 723         unsigned no_fsync:1;
 724
 725         /** Is fsyncdir not implemented by fs? */
 726         unsigned no_fsyncdir:1;
 727
 728         /** Is flush not implemented by fs? */
 729         unsigned no_flush:1;
 730
 731         /** Is setxattr not implemented by fs? */
 732         unsigned no_setxattr:1;
 733
 734         /** Does file server support extended setxattr */
 735         unsigned setxattr_ext:1;
 736
 737         /** Is getxattr not implemented by fs? */
 738         unsigned no_getxattr:1;
 739
 740         /** Is listxattr not implemented by fs? */
 741         unsigned no_listxattr:1;
 742
 743         /** Is removexattr not implemented by fs? */
 744         unsigned no_removexattr:1;
 745
 746         /** Are posix file locking primitives not implemented by fs? */
 747         unsigned no_lock:1;
 748
 749         /** Is access not implemented by fs? */
 750         unsigned no_access:1;
 751
 752         /** Is create not implemented by fs? */
 753         unsigned no_create:1;
 754
 755         /** Is interrupt not implemented by fs? */
 756         unsigned no_interrupt:1;
 757
 758         /** Is bmap not implemented by fs? */
 759         unsigned no_bmap:1;
 760
 761         /** Is poll not implemented by fs? */
 762         unsigned no_poll:1;
 763
 764         /** Do multi-page cached writes */
 765         unsigned big_writes:1;
 766
 767         /** Don't apply umask to creation modes */
 768         unsigned dont_mask:1;
 769
 770         /** Are BSD file locking primitives not implemented by fs? */
 771         unsigned no_flock:1;
 772
 773         /** Is fallocate not implemented by fs? */
 774         unsigned no_fallocate:1;
 775
 776         /** Is rename with flags implemented by fs? */
 777         unsigned no_rename2:1;
 778
 779         /** Use enhanced/automatic page cache invalidation. */
 780         unsigned auto_inval_data:1;
 781
 782         /** Filesystem is fully responsible for page cache invalidation. */
 783         unsigned explicit_inval_data:1;
 784
 785         /** Does the filesystem support readdirplus? */
 786         unsigned do_readdirplus:1;
 787
 788         /** Does the filesystem want adaptive readdirplus? */
 789         unsigned readdirplus_auto:1;
 790
 791         /** Does the filesystem support asynchronous direct-IO submission? */
 792         unsigned async_dio:1;
 793
 794         /** Is lseek not implemented by fs? */
 795         unsigned no_lseek:1;
 796
 797         /** Does the filesystem support posix acls? */
 798         unsigned posix_acl:1;
 799
 800         /** Check permissions based on the file mode or not? */
 801         unsigned default_permissions:1;
 802
 803         /** Allow other than the mounter user to access the filesystem ? */
 804         unsigned allow_other:1;
 805
 806         /** Does the filesystem support copy_file_range? */
 807         unsigned no_copy_file_range:1;
 808
 809         /* Send DESTROY request */
 810         unsigned int destroy:1;
 811
 812         /* Delete dentries that have gone stale */
 813         unsigned int delete_stale:1;
 814
 815         /** Do not create entry in fusectl fs */
 816         unsigned int no_control:1;
 817
 818         /** Do not allow MNT_FORCE umount */
 819         unsigned int no_force_umount:1;
 820
 821         /* Auto-mount submounts announced by the server */
 822         unsigned int auto_submounts:1;
 823
 824         /* Propagate syncfs() to server */
 825         unsigned int sync_fs:1;
 826
 827         /* Initialize security xattrs when creating a new inode */
 828         unsigned int init_security:1;
 829
 830         /* Add supplementary group info when creating a new inode */
 831         unsigned int create_supp_group:1;
 832
 833         /* Does the filesystem support per inode DAX? */
 834         unsigned int inode_dax:1;
 835
 836         /* Is tmpfile not implemented by fs? */
 837         unsigned int no_tmpfile:1;
 838
 839         /* Relax restrictions to allow shared mmap in FOPEN_DIRECT_IO mode */
 840         unsigned int direct_io_allow_mmap:1;
 841
 842         /* Is statx not implemented by fs? */
 843         unsigned int no_statx:1;
 844
 845         /** Passthrough support for read/write IO */
 846         unsigned int passthrough:1;
 847
 848         /** Maximum stack depth for passthrough backing files */
 849         int max_stack_depth;
 850
 851         /** The number of requests waiting for completion */
 852         atomic_t num_waiting;
 853
 854         /** Negotiated minor version */
 855         unsigned minor;
 856
 857         /** Entry on the fuse_mount_list */
 858         struct list_head entry;
 859
 860         /** Device ID from the root super block */
 861         dev_t dev;
 862
 863         /** Dentries in the control filesystem */
 864         struct dentry *ctl_dentry[FUSE_CTL_NUM_DENTRIES];
 865
 866         /** number of dentries used in the above array */
 867         int ctl_ndents;
 868
 869         /** Key for lock owner ID scrambling */
 870         u32 scramble_key[4];
 871
 872         /** Version counter for attribute changes */
 873         atomic64_t attr_version;
 874
 875         /** Called on final put */
 876         void (*release)(struct fuse_conn *);
 877
 878         /**
 879          * Read/write semaphore to hold when accessing the sb of any
 880          * fuse_mount belonging to this connection
 881          */
 882         struct rw_semaphore killsb;
 883
 884         /** List of device instances belonging to this connection */
 885         struct list_head devices;
 886
 887 #ifdef CONFIG_FUSE_DAX
 888         /* Dax mode */
 889         enum fuse_dax_mode dax_mode;
 890
 891         /* Dax specific conn data, non-NULL if DAX is enabled */
 892         struct fuse_conn_dax *dax;
 893 #endif
 894
 895         /** List of filesystems using this connection */
 896         struct list_head mounts;
 897
 898         /* New writepages go into this bucket */
 899         struct fuse_sync_bucket __rcu *curr_bucket;
 900 };
 901
 902 /*
 903  * Represents a mounted filesystem, potentially a submount.
 904  *
 905  * This object allows sharing a fuse_conn between separate mounts to
 906  * allow submounts with dedicated superblocks and thus separate device
 907  * IDs.
 908  */
 909 struct fuse_mount {
 910         /* Underlying (potentially shared) connection to the FUSE server */
 911         struct fuse_conn *fc;
 912
 913         /*
 914          * Super block for this connection (fc->killsb must be held when
 915          * accessing this).
 916          */
 917         struct super_block *sb;
 918
 919         /* Entry on fc->mounts */
 920         struct list_head fc_entry;
 921 };
 922
 923 static inline struct fuse_mount *get_fuse_mount_super(struct super_block *sb)
 924 {
 925         return sb->s_fs_info;
 926 }
 927
 928 static inline struct fuse_conn *get_fuse_conn_super(struct super_block *sb)
 929 {
 930         return get_fuse_mount_super(sb)->fc;
 931 }
 932
 933 static inline struct fuse_mount *get_fuse_mount(struct inode *inode)
 934 {
 935         return get_fuse_mount_super(inode->i_sb);
 936 }
 937
 938 static inline struct fuse_conn *get_fuse_conn(struct inode *inode)
 939 {
 940         return get_fuse_mount_super(inode->i_sb)->fc;
 941 }
 942
 943 static inline struct fuse_inode *get_fuse_inode(struct inode *inode)
 944 {
 945         return container_of(inode, struct fuse_inode, inode);
 946 }
 947
 948 static inline u64 get_node_id(struct inode *inode)
 949 {
 950         return get_fuse_inode(inode)->nodeid;
 951 }
 952
 953 static inline int invalid_nodeid(u64 nodeid)
 954 {
 955         return !nodeid || nodeid == FUSE_ROOT_ID;
 956 }
 957
 958 static inline u64 fuse_get_attr_version(struct fuse_conn *fc)
 959 {
 960         return atomic64_read(&fc->attr_version);
 961 }
 962
 963 static inline bool fuse_stale_inode(const struct inode *inode, int generation,
 964                                     struct fuse_attr *attr)
 965 {
 966         return inode->i_generation != generation ||
 967                 inode_wrong_type(inode, attr->mode);
 968 }
 969
 970 static inline void fuse_make_bad(struct inode *inode)
 971 {
 972         remove_inode_hash(inode);
 973         set_bit(FUSE_I_BAD, &get_fuse_inode(inode)->state);
 974 }
 975
 976 static inline bool fuse_is_bad(struct inode *inode)
 977 {
 978         return unlikely(test_bit(FUSE_I_BAD, &get_fuse_inode(inode)->state));
 979 }
 980
 981 static inline struct page **fuse_pages_alloc(unsigned int npages, gfp_t flags,
 982                                              struct fuse_page_desc **desc)
 983 {
 984         struct page **pages;
 985
 986         pages = kzalloc(npages * (sizeof(struct page *) +
 987                                   sizeof(struct fuse_page_desc)), flags);
 988         *desc = (void *) (pages + npages);
 989
 990         return pages;
 991 }
 992
 993 static inline void fuse_page_descs_length_init(struct fuse_page_desc *descs,
 994                                                unsigned int index,
 995                                                unsigned int nr_pages)
 996 {
 997         int i;
 998
 999         for (i = index; i < index + nr_pages; i++)
1000                 descs[i].length = PAGE_SIZE - descs[i].offset;
1001 }
1002
1003 static inline void fuse_sync_bucket_dec(struct fuse_sync_bucket *bucket)
1004 {
1005         /* Need RCU protection to prevent use after free after the decrement */
1006         rcu_read_lock();
1007         if (atomic_dec_and_test(&bucket->count))
1008                 wake_up(&bucket->waitq);
1009         rcu_read_unlock();
1010 }
1011
1012 /** Device operations */
1013 extern const struct file_operations fuse_dev_operations;
1014
1015 extern const struct dentry_operations fuse_dentry_operations;
1016 extern const struct dentry_operations fuse_root_dentry_operations;
1017
1018 /**
1019  * Get a filled in inode
1020  */
1021 struct inode *fuse_iget(struct super_block *sb, u64 nodeid,
1022                         int generation, struct fuse_attr *attr,
1023                         u64 attr_valid, u64 attr_version);
1024
1025 int fuse_lookup_name(struct super_block *sb, u64 nodeid, const struct qstr *name,
1026                      struct fuse_entry_out *outarg, struct inode **inode);
1027
1028 /**
1029  * Send FORGET command
1030  */
1031 void fuse_queue_forget(struct fuse_conn *fc, struct fuse_forget_link *forget,
1032                        u64 nodeid, u64 nlookup);
1033
1034 struct fuse_forget_link *fuse_alloc_forget(void);
1035
1036 struct fuse_forget_link *fuse_dequeue_forget(struct fuse_iqueue *fiq,
1037                                              unsigned int max,
1038                                              unsigned int *countp);
1039
1040 /*
1041  * Initialize READ or READDIR request
1042  */
1043 struct fuse_io_args {
1044         union {
1045                 struct {
1046                         struct fuse_read_in in;
1047                         u64 attr_ver;
1048                 } read;
1049                 struct {
1050                         struct fuse_write_in in;
1051                         struct fuse_write_out out;
1052                         bool page_locked;
1053                 } write;
1054         };
1055         struct fuse_args_pages ap;
1056         struct fuse_io_priv *io;
1057         struct fuse_file *ff;
1058 };
1059
1060 void fuse_read_args_fill(struct fuse_io_args *ia, struct file *file, loff_t pos,
1061                          size_t count, int opcode);
1062
1063
1064 struct fuse_file *fuse_file_alloc(struct fuse_mount *fm, bool release);
1065 void fuse_file_free(struct fuse_file *ff);
1066 int fuse_finish_open(struct inode *inode, struct file *file);
1067
1068 void fuse_sync_release(struct fuse_inode *fi, struct fuse_file *ff,
1069                        unsigned int flags);
1070
1071 /**
1072  * Send RELEASE or RELEASEDIR request
1073  */
1074 void fuse_release_common(struct file *file, bool isdir);
1075
1076 /**
1077  * Send FSYNC or FSYNCDIR request
1078  */
1079 int fuse_fsync_common(struct file *file, loff_t start, loff_t end,
1080                       int datasync, int opcode);
1081
1082 /**
1083  * Notify poll wakeup
1084  */
1085 int fuse_notify_poll_wakeup(struct fuse_conn *fc,
1086                             struct fuse_notify_poll_wakeup_out *outarg);
1087
1088 /**
1089  * Initialize file operations on a regular file
1090  */
1091 void fuse_init_file_inode(struct inode *inode, unsigned int flags);
1092
1093 /**
1094  * Initialize inode operations on regular files and special files
1095  */
1096 void fuse_init_common(struct inode *inode);
1097
1098 /**
1099  * Initialize inode and file operations on a directory
1100  */
1101 void fuse_init_dir(struct inode *inode);
1102
1103 /**
1104  * Initialize inode operations on a symlink
1105  */
1106 void fuse_init_symlink(struct inode *inode);
1107
1108 /**
1109  * Change attributes of an inode
1110  */
1111 void fuse_change_attributes(struct inode *inode, struct fuse_attr *attr,
1112                             struct fuse_statx *sx,
1113                             u64 attr_valid, u64 attr_version);
1114
1115 void fuse_change_attributes_common(struct inode *inode, struct fuse_attr *attr,
1116                                    struct fuse_statx *sx,
1117                                    u64 attr_valid, u32 cache_mask);
1118
1119 u32 fuse_get_cache_mask(struct inode *inode);
1120
1121 /**
1122  * Initialize the client device
1123  */
1124 int fuse_dev_init(void);
1125
1126 /**
1127  * Cleanup the client device
1128  */
1129 void fuse_dev_cleanup(void);
1130
1131 int fuse_ctl_init(void);
1132 void __exit fuse_ctl_cleanup(void);
1133
1134 /**
1135  * Simple request sending that does request allocation and freeing
1136  */
1137 ssize_t fuse_simple_request(struct fuse_mount *fm, struct fuse_args *args);
1138 int fuse_simple_background(struct fuse_mount *fm, struct fuse_args *args,
1139                            gfp_t gfp_flags);
1140
1141 /**
1142  * End a finished request
1143  */
1144 void fuse_request_end(struct fuse_req *req);
1145
1146 /* Abort all requests */
1147 void fuse_abort_conn(struct fuse_conn *fc);
1148 void fuse_wait_aborted(struct fuse_conn *fc);
1149
1150 /**
1151  * Invalidate inode attributes
1152  */
1153
1154 /* Attributes possibly changed on data modification */
1155 #define FUSE_STATX_MODIFY       (STATX_MTIME | STATX_CTIME | STATX_BLOCKS)
1156
1157 /* Attributes possibly changed on data and/or size modification */
1158 #define FUSE_STATX_MODSIZE      (FUSE_STATX_MODIFY | STATX_SIZE)
1159
1160 void fuse_invalidate_attr(struct inode *inode);
1161 void fuse_invalidate_attr_mask(struct inode *inode, u32 mask);
1162
1163 void fuse_invalidate_entry_cache(struct dentry *entry);
1164
1165 void fuse_invalidate_atime(struct inode *inode);
1166
1167 u64 fuse_time_to_jiffies(u64 sec, u32 nsec);
1168 #define ATTR_TIMEOUT(o) \
1169         fuse_time_to_jiffies((o)->attr_valid, (o)->attr_valid_nsec)
1170
1171 void fuse_change_entry_timeout(struct dentry *entry, struct fuse_entry_out *o);
1172
1173 /**
1174  * Acquire reference to fuse_conn
1175  */
1176 struct fuse_conn *fuse_conn_get(struct fuse_conn *fc);
1177
1178 /**
1179  * Initialize fuse_conn
1180  */
1181 void fuse_conn_init(struct fuse_conn *fc, struct fuse_mount *fm,
1182                     struct user_namespace *user_ns,
1183                     const struct fuse_iqueue_ops *fiq_ops, void *fiq_priv);
1184
1185 /**
1186  * Release reference to fuse_conn
1187  */
1188 void fuse_conn_put(struct fuse_conn *fc);
1189
1190 struct fuse_dev *fuse_dev_alloc_install(struct fuse_conn *fc);
1191 struct fuse_dev *fuse_dev_alloc(void);
1192 void fuse_dev_install(struct fuse_dev *fud, struct fuse_conn *fc);
1193 void fuse_dev_free(struct fuse_dev *fud);
1194 void fuse_send_init(struct fuse_mount *fm);
1195
1196 /**
1197  * Fill in superblock and initialize fuse connection
1198  * @sb: partially-initialized superblock to fill in
1199  * @ctx: mount context
1200  */
1201 int fuse_fill_super_common(struct super_block *sb, struct fuse_fs_context *ctx);
1202
1203 /*
1204  * Remove the mount from the connection
1205  *
1206  * Returns whether this was the last mount
1207  */
1208 bool fuse_mount_remove(struct fuse_mount *fm);
1209
1210 /*
1211  * Setup context ops for submounts
1212  */
1213 int fuse_init_fs_context_submount(struct fs_context *fsc);
1214
1215 /*
1216  * Shut down the connection (possibly sending DESTROY request).
1217  */
1218 void fuse_conn_destroy(struct fuse_mount *fm);
1219
1220 /* Drop the connection and free the fuse mount */
1221 void fuse_mount_destroy(struct fuse_mount *fm);
1222
1223 /**
1224  * Add connection to control filesystem
1225  */
1226 int fuse_ctl_add_conn(struct fuse_conn *fc);
1227
1228 /**
1229  * Remove connection from control filesystem
1230  */
1231 void fuse_ctl_remove_conn(struct fuse_conn *fc);
1232
1233 /**
1234  * Is file type valid?
1235  */
1236 int fuse_valid_type(int m);
1237
1238 bool fuse_invalid_attr(struct fuse_attr *attr);
1239
1240 /**
1241  * Is current process allowed to perform filesystem operation?
1242  */
1243 bool fuse_allow_current_process(struct fuse_conn *fc);
1244
1245 u64 fuse_lock_owner_id(struct fuse_conn *fc, fl_owner_t id);
1246
1247 void fuse_flush_time_update(struct inode *inode);
1248 void fuse_update_ctime(struct inode *inode);
1249
1250 int fuse_update_attributes(struct inode *inode, struct file *file, u32 mask);
1251
1252 void fuse_flush_writepages(struct inode *inode);
1253
1254 void fuse_set_nowrite(struct inode *inode);
1255 void fuse_release_nowrite(struct inode *inode);
1256
1257 /**
1258  * Scan all fuse_mounts belonging to fc to find the first where
1259  * ilookup5() returns a result.  Return that result and the
1260  * respective fuse_mount in *fm (unless fm is NULL).
1261  *
1262  * The caller must hold fc->killsb.
1263  */
1264 struct inode *fuse_ilookup(struct fuse_conn *fc, u64 nodeid,
1265                            struct fuse_mount **fm);
1266
1267 /**
1268  * File-system tells the kernel to invalidate cache for the given node id.
1269  */
1270 int fuse_reverse_inval_inode(struct fuse_conn *fc, u64 nodeid,
1271                              loff_t offset, loff_t len);
1272
1273 /**
1274  * File-system tells the kernel to invalidate parent attributes and
1275  * the dentry matching parent/name.
1276  *
1277  * If the child_nodeid is non-zero and:
1278  *    - matches the inode number for the dentry matching parent/name,
1279  *    - is not a mount point
1280  *    - is a file or oan empty directory
1281  * then the dentry is unhashed (d_delete()).
1282  */
1283 int fuse_reverse_inval_entry(struct fuse_conn *fc, u64 parent_nodeid,
1284                              u64 child_nodeid, struct qstr *name, u32 flags);
1285
1286 int fuse_do_open(struct fuse_mount *fm, u64 nodeid, struct file *file,
1287                  bool isdir);
1288
1289 /**
1290  * fuse_direct_io() flags
1291  */
1292
1293 /** If set, it is WRITE; otherwise - READ */
1294 #define FUSE_DIO_WRITE (1 << 0)
1295
1296 /** CUSE pass fuse_direct_io() a file which f_mapping->host is not from FUSE */
1297 #define FUSE_DIO_CUSE  (1 << 1)
1298
1299 ssize_t fuse_direct_io(struct fuse_io_priv *io, struct iov_iter *iter,
1300                        loff_t *ppos, int flags);
1301 long fuse_do_ioctl(struct file *file, unsigned int cmd, unsigned long arg,
1302                    unsigned int flags);
1303 long fuse_ioctl_common(struct file *file, unsigned int cmd,
1304                        unsigned long arg, unsigned int flags);
1305 __poll_t fuse_file_poll(struct file *file, poll_table *wait);
1306 int fuse_dev_release(struct inode *inode, struct file *file);
1307
1308 bool fuse_write_update_attr(struct inode *inode, loff_t pos, ssize_t written);
1309
1310 int fuse_flush_times(struct inode *inode, struct fuse_file *ff);
1311 int fuse_write_inode(struct inode *inode, struct writeback_control *wbc);
1312
1313 int fuse_do_setattr(struct dentry *dentry, struct iattr *attr,
1314                     struct file *file);
1315
1316 void fuse_set_initialized(struct fuse_conn *fc);
1317
1318 void fuse_unlock_inode(struct inode *inode, bool locked);
1319 bool fuse_lock_inode(struct inode *inode);
1320
1321 int fuse_setxattr(struct inode *inode, const char *name, const void *value,
1322                   size_t size, int flags, unsigned int extra_flags);
1323 ssize_t fuse_getxattr(struct inode *inode, const char *name, void *value,
1324                       size_t size);
1325 ssize_t fuse_listxattr(struct dentry *entry, char *list, size_t size);
1326 int fuse_removexattr(struct inode *inode, const char *name);
1327 extern const struct xattr_handler * const fuse_xattr_handlers[];
1328
1329 struct posix_acl;
1330 struct posix_acl *fuse_get_inode_acl(struct inode *inode, int type, bool rcu);
1331 struct posix_acl *fuse_get_acl(struct mnt_idmap *idmap,
1332                                struct dentry *dentry, int type);
1333 int fuse_set_acl(struct mnt_idmap *, struct dentry *dentry,
1334                  struct posix_acl *acl, int type);
1335
1336 /* readdir.c */
1337 int fuse_readdir(struct file *file, struct dir_context *ctx);
1338
1339 /**
1340  * Return the number of bytes in an arguments list
1341  */
1342 unsigned int fuse_len_args(unsigned int numargs, struct fuse_arg *args);
1343
1344 /**
1345  * Get the next unique ID for a request
1346  */
1347 u64 fuse_get_unique(struct fuse_iqueue *fiq);
1348 void fuse_free_conn(struct fuse_conn *fc);
1349
1350 /* dax.c */
1351
1352 #define FUSE_IS_DAX(inode) (IS_ENABLED(CONFIG_FUSE_DAX) && IS_DAX(inode))
1353
1354 ssize_t fuse_dax_read_iter(struct kiocb *iocb, struct iov_iter *to);
1355 ssize_t fuse_dax_write_iter(struct kiocb *iocb, struct iov_iter *from);
1356 int fuse_dax_mmap(struct file *file, struct vm_area_struct *vma);
1357 int fuse_dax_break_layouts(struct inode *inode, u64 dmap_start, u64 dmap_end);
1358 int fuse_dax_conn_alloc(struct fuse_conn *fc, enum fuse_dax_mode mode,
1359                         struct dax_device *dax_dev);
1360 void fuse_dax_conn_free(struct fuse_conn *fc);
1361 bool fuse_dax_inode_alloc(struct super_block *sb, struct fuse_inode *fi);
1362 void fuse_dax_inode_init(struct inode *inode, unsigned int flags);
1363 void fuse_dax_inode_cleanup(struct inode *inode);
1364 void fuse_dax_dontcache(struct inode *inode, unsigned int flags);
1365 bool fuse_dax_check_alignment(struct fuse_conn *fc, unsigned int map_alignment);
1366 void fuse_dax_cancel_work(struct fuse_conn *fc);
1367
1368 /* ioctl.c */
1369 long fuse_file_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
1370 long fuse_file_compat_ioctl(struct file *file, unsigned int cmd,
1371                             unsigned long arg);
1372 int fuse_fileattr_get(struct dentry *dentry, struct fileattr *fa);
1373 int fuse_fileattr_set(struct mnt_idmap *idmap,
1374                       struct dentry *dentry, struct fileattr *fa);
1375
1376 /* iomode.c */
1377 int fuse_file_cached_io_start(struct inode *inode, struct fuse_file *ff);
1378 int fuse_file_uncached_io_start(struct inode *inode, struct fuse_file *ff);
1379 void fuse_file_uncached_io_end(struct inode *inode, struct fuse_file *ff);
1380
1381 int fuse_file_io_open(struct file *file, struct inode *inode);
1382 void fuse_file_io_release(struct fuse_file *ff, struct inode *inode);
1383
1384 /* file.c */
1385 struct fuse_file *fuse_file_open(struct fuse_mount *fm, u64 nodeid,
1386                                  unsigned int open_flags, bool isdir);
1387 void fuse_file_release(struct inode *inode, struct fuse_file *ff,
1388                        unsigned int open_flags, fl_owner_t id, bool isdir);
1389
1390 /* passthrough.c */
1391 static inline struct fuse_backing *fuse_inode_backing(struct fuse_inode *fi)
1392 {
1393 #ifdef CONFIG_FUSE_PASSTHROUGH
1394         return READ_ONCE(fi->fb);
1395 #else
1396         return NULL;
1397 #endif
1398 }
1399
1400 static inline struct fuse_backing *fuse_inode_backing_set(struct fuse_inode *fi,
1401                                                           struct fuse_backing *fb)
1402 {
1403 #ifdef CONFIG_FUSE_PASSTHROUGH
1404         return xchg(&fi->fb, fb);
1405 #else
1406         return NULL;
1407 #endif
1408 }
1409
1410 struct fuse_backing *fuse_backing_get(struct fuse_backing *fb);
1411 void fuse_backing_put(struct fuse_backing *fb);
1412
1413 #endif /* _FS_FUSE_I_H */