super.h

   1/* SPDX-License-Identifier: GPL-2.0 */
   2#ifndef _FS_CEPH_SUPER_H
   3#define _FS_CEPH_SUPER_H
   4
   5#include <linux/ceph/ceph_debug.h>
   6#include <linux/ceph/osd_client.h>
   7
   8#include <linux/unaligned.h>
   9#include <linux/backing-dev.h>
  10#include <linux/completion.h>
  11#include <linux/exportfs.h>
  12#include <linux/fs.h>
  13#include <linux/mempool.h>
  14#include <linux/pagemap.h>
  15#include <linux/wait.h>
  16#include <linux/writeback.h>
  17#include <linux/slab.h>
  18#include <linux/posix_acl.h>
  19#include <linux/refcount.h>
  20#include <linux/security.h>
  21#include <linux/netfs.h>
  22#include <linux/fscache.h>
  23#include <linux/hashtable.h>
  24
  25#include <linux/ceph/libceph.h>
  26#include "crypto.h"
  27
  28/* large granularity for statfs utilization stats to facilitate
  29 * large volume sizes on 32-bit machines. */
  30#define CEPH_BLOCK_SHIFT   22  /* 4 MB */
  31#define CEPH_BLOCK         (1 << CEPH_BLOCK_SHIFT)
  32#define CEPH_4K_BLOCK_SHIFT 12  /* 4 KB */
  33
  34#define CEPH_MOUNT_OPT_CLEANRECOVER    (1<<1) /* auto reonnect (clean mode) after blocklisted */
  35#define CEPH_MOUNT_OPT_DIRSTAT         (1<<4) /* `cat dirname` for stats */
  36#define CEPH_MOUNT_OPT_RBYTES          (1<<5) /* dir st_bytes = rbytes */
  37#define CEPH_MOUNT_OPT_NOASYNCREADDIR  (1<<7) /* no dcache readdir */
  38#define CEPH_MOUNT_OPT_INO32           (1<<8) /* 32 bit inos */
  39#define CEPH_MOUNT_OPT_DCACHE          (1<<9) /* use dcache for readdir etc */
  40#define CEPH_MOUNT_OPT_FSCACHE         (1<<10) /* use fscache */
  41#define CEPH_MOUNT_OPT_NOPOOLPERM      (1<<11) /* no pool permission check */
  42#define CEPH_MOUNT_OPT_MOUNTWAIT       (1<<12) /* mount waits if no mds is up */
  43#define CEPH_MOUNT_OPT_NOQUOTADF       (1<<13) /* no root dir quota in statfs */
  44#define CEPH_MOUNT_OPT_NOCOPYFROM      (1<<14) /* don't use RADOS 'copy-from' op */
  45#define CEPH_MOUNT_OPT_ASYNC_DIROPS    (1<<15) /* allow async directory ops */
  46#define CEPH_MOUNT_OPT_NOPAGECACHE     (1<<16) /* bypass pagecache altogether */
  47#define CEPH_MOUNT_OPT_SPARSEREAD      (1<<17) /* always do sparse reads */
  48
  49#define CEPH_MOUNT_OPT_DEFAULT			\
  50	(CEPH_MOUNT_OPT_DCACHE |		\
  51	 CEPH_MOUNT_OPT_NOCOPYFROM |		\
  52	 CEPH_MOUNT_OPT_ASYNC_DIROPS)
  53
  54#define ceph_set_mount_opt(fsc, opt) \
  55	(fsc)->mount_options->flags |= CEPH_MOUNT_OPT_##opt
  56#define ceph_clear_mount_opt(fsc, opt) \
  57	(fsc)->mount_options->flags &= ~CEPH_MOUNT_OPT_##opt
  58#define ceph_test_mount_opt(fsc, opt) \
  59	(!!((fsc)->mount_options->flags & CEPH_MOUNT_OPT_##opt))
  60
  61/* max size of osd read request, limited by libceph */
  62#define CEPH_MAX_READ_SIZE              CEPH_MSG_MAX_DATA_LEN
  63/* osd has a configurable limitation of max write size.
  64 * CEPH_MSG_MAX_DATA_LEN should be small enough. */
  65#define CEPH_MAX_WRITE_SIZE		CEPH_MSG_MAX_DATA_LEN
  66#define CEPH_RASIZE_DEFAULT             (8192*1024)    /* max readahead */
  67#define CEPH_MAX_READDIR_DEFAULT        1024
  68#define CEPH_MAX_READDIR_BYTES_DEFAULT  (512*1024)
  69#define CEPH_SNAPDIRNAME_DEFAULT        ".snap"
  70
  71/*
  72 * Delay telling the MDS we no longer want caps, in case we reopen
  73 * the file.  Delay a minimum amount of time, even if we send a cap
  74 * message for some other reason.  Otherwise, take the oppotunity to
  75 * update the mds to avoid sending another message later.
  76 */
  77#define CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT      5  /* cap release delay */
  78#define CEPH_CAPS_WANTED_DELAY_MAX_DEFAULT     60  /* cap release delay */
  79
  80struct ceph_mount_options {
  81	unsigned int flags;
  82
  83	unsigned int wsize;            /* max write size */
  84	unsigned int rsize;            /* max read size */
  85	unsigned int rasize;           /* max readahead */
  86	unsigned int congestion_kb;    /* max writeback in flight */
  87	unsigned int caps_wanted_delay_min, caps_wanted_delay_max;
  88	int caps_max;
  89	unsigned int max_readdir;       /* max readdir result (entries) */
  90	unsigned int max_readdir_bytes; /* max readdir result (bytes) */
  91
  92	bool new_dev_syntax;
  93
  94	/*
  95	 * everything above this point can be memcmp'd; everything below
  96	 * is handled in compare_mount_options()
  97	 */
  98
  99	char *snapdir_name;   /* default ".snap" */
 100	char *mds_namespace;  /* default NULL */
 101	char *server_path;    /* default NULL (means "/") */
 102	char *fscache_uniq;   /* default NULL */
 103	char *mon_addr;
 104	struct fscrypt_dummy_policy dummy_enc_policy;
 105};
 106
 107#define CEPH_NAMESPACE_WILDCARD		"*"
 108
 109static inline bool ceph_namespace_match(const char *pattern,
 110					const char *target)
 111{
 112	if (!pattern || !pattern[0] ||
 113	    !strcmp(pattern, CEPH_NAMESPACE_WILDCARD))
 114		return true;
 115
 116	return !strcmp(pattern, target);
 117}
 118
 119/*
 120 * Check if the mds namespace in ceph_mount_options matches
 121 * the passed in namespace string. First time match (when
 122 * ->mds_namespace is NULL) is treated specially, since
 123 * ->mds_namespace needs to be initialized by the caller.
 124 */
 125static inline bool namespace_equals(struct ceph_mount_options *fsopt,
 126				    const char *namespace, size_t len)
 127{
 128	return !(fsopt->mds_namespace &&
 129		 (strlen(fsopt->mds_namespace) != len ||
 130		  strncmp(fsopt->mds_namespace, namespace, len)));
 131}
 132
 133/* mount state */
 134enum {
 135	CEPH_MOUNT_MOUNTING,
 136	CEPH_MOUNT_MOUNTED,
 137	CEPH_MOUNT_UNMOUNTING,
 138	CEPH_MOUNT_UNMOUNTED,
 139	CEPH_MOUNT_SHUTDOWN,
 140	CEPH_MOUNT_RECOVER,
 141	CEPH_MOUNT_FENCE_IO,
 142};
 143
 144#define CEPH_ASYNC_CREATE_CONFLICT_BITS 8
 145
 146struct ceph_fs_client {
 147	struct super_block *sb;
 148
 149	struct list_head metric_wakeup;
 150
 151	struct ceph_mount_options *mount_options;
 152	struct ceph_client *client;
 153
 154	int mount_state;
 155
 156	bool blocklisted;
 157
 158	bool have_copy_from2;
 159
 160	u32 filp_gen;
 161	loff_t max_file_size;
 162
 163	struct ceph_mds_client *mdsc;
 164
 165	atomic_long_t writeback_count;
 166	bool write_congested;
 167
 168	struct workqueue_struct *inode_wq;
 169	struct workqueue_struct *cap_wq;
 170
 171	DECLARE_HASHTABLE(async_unlink_conflict, CEPH_ASYNC_CREATE_CONFLICT_BITS);
 172	spinlock_t async_unlink_conflict_lock;
 173
 174#ifdef CONFIG_DEBUG_FS
 175	struct dentry *debugfs_dentry_lru, *debugfs_caps;
 176	struct dentry *debugfs_congestion_kb;
 177	struct dentry *debugfs_bdi;
 178	struct dentry *debugfs_mdsc, *debugfs_mdsmap;
 179	struct dentry *debugfs_status;
 180	struct dentry *debugfs_mds_sessions;
 181	struct dentry *debugfs_metrics_dir;
 182#endif
 183
 184#ifdef CONFIG_CEPH_FSCACHE
 185	struct fscache_volume *fscache;
 186#endif
 187#ifdef CONFIG_FS_ENCRYPTION
 188	struct fscrypt_dummy_policy fsc_dummy_enc_policy;
 189#endif
 190};
 191
 192/*
 193 * File i/o capability.  This tracks shared state with the metadata
 194 * server that allows us to cache or writeback attributes or to read
 195 * and write data.  For any given inode, we should have one or more
 196 * capabilities, one issued by each metadata server, and our
 197 * cumulative access is the OR of all issued capabilities.
 198 *
 199 * Each cap is referenced by the inode's i_caps rbtree and by per-mds
 200 * session capability lists.
 201 */
 202struct ceph_cap {
 203	struct ceph_inode_info *ci;
 204	struct rb_node ci_node;          /* per-ci cap tree */
 205	struct ceph_mds_session *session;
 206	struct list_head session_caps;   /* per-session caplist */
 207	u64 cap_id;       /* unique cap id (mds provided) */
 208	union {
 209		/* in-use caps */
 210		struct {
 211			int issued;       /* latest, from the mds */
 212			int implemented;  /* implemented superset of
 213					     issued (for revocation) */
 214			int mds;	  /* mds index for this cap */
 215			int mds_wanted;   /* caps wanted from this mds */
 216		};
 217		/* caps to release */
 218		struct {
 219			u64 cap_ino;
 220			int queue_release;
 221		};
 222	};
 223	u32 seq, issue_seq, mseq;
 224	u32 cap_gen;      /* active/stale cycle */
 225	unsigned long last_used;
 226	struct list_head caps_item;
 227};
 228
 229#define CHECK_CAPS_AUTHONLY     1  /* only check auth cap */
 230#define CHECK_CAPS_FLUSH        2  /* flush any dirty caps */
 231#define CHECK_CAPS_NOINVAL      4  /* don't invalidate pagecache */
 232#define CHECK_CAPS_FLUSH_FORCE  8  /* force flush any caps */
 233
 234struct ceph_cap_flush {
 235	u64 tid;
 236	int caps;
 237	bool wake; /* wake up flush waiters when finish ? */
 238	bool is_capsnap; /* true means capsnap */
 239	struct list_head g_list; // global
 240	struct list_head i_list; // per inode
 241};
 242
 243/*
 244 * Snapped cap state that is pending flush to mds.  When a snapshot occurs,
 245 * we first complete any in-process sync writes and writeback any dirty
 246 * data before flushing the snapped state (tracked here) back to the MDS.
 247 */
 248struct ceph_cap_snap {
 249	refcount_t nref;
 250	struct list_head ci_item;
 251
 252	struct ceph_cap_flush cap_flush;
 253
 254	u64 follows;
 255	int issued, dirty;
 256	struct ceph_snap_context *context;
 257
 258	umode_t mode;
 259	kuid_t uid;
 260	kgid_t gid;
 261
 262	struct ceph_buffer *xattr_blob;
 263	u64 xattr_version;
 264
 265	u64 size;
 266	u64 change_attr;
 267	struct timespec64 mtime, atime, ctime, btime;
 268	u64 time_warp_seq;
 269	u64 truncate_size;
 270	u32 truncate_seq;
 271	int writing;   /* a sync write is still in progress */
 272	int dirty_pages;     /* dirty pages awaiting writeback */
 273	bool inline_data;
 274	bool need_flush;
 275};
 276
 277static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap)
 278{
 279	if (refcount_dec_and_test(&capsnap->nref)) {
 280		if (capsnap->xattr_blob)
 281			ceph_buffer_put(capsnap->xattr_blob);
 282		kmem_cache_free(ceph_cap_snap_cachep, capsnap);
 283	}
 284}
 285
 286/*
 287 * The frag tree describes how a directory is fragmented, potentially across
 288 * multiple metadata servers.  It is also used to indicate points where
 289 * metadata authority is delegated, and whether/where metadata is replicated.
 290 *
 291 * A _leaf_ frag will be present in the i_fragtree IFF there is
 292 * delegation info.  That is, if mds >= 0 || ndist > 0.
 293 */
 294#define CEPH_MAX_DIRFRAG_REP 4
 295
 296struct ceph_inode_frag {
 297	struct rb_node node;
 298
 299	/* fragtree state */
 300	u32 frag;
 301	int split_by;         /* i.e. 2^(split_by) children */
 302
 303	/* delegation and replication info */
 304	int mds;              /* -1 if same authority as parent */
 305	int ndist;            /* >0 if replicated */
 306	int dist[CEPH_MAX_DIRFRAG_REP];
 307};
 308
 309/*
 310 * We cache inode xattrs as an encoded blob until they are first used,
 311 * at which point we parse them into an rbtree.
 312 */
 313struct ceph_inode_xattr {
 314	struct rb_node node;
 315
 316	const char *name;
 317	int name_len;
 318	const char *val;
 319	int val_len;
 320	int dirty;
 321
 322	int should_free_name;
 323	int should_free_val;
 324};
 325
 326/*
 327 * Ceph dentry state
 328 */
 329struct ceph_dentry_info {
 330	struct dentry *dentry;
 331	struct ceph_mds_session *lease_session;
 332	struct list_head lease_list;
 333	struct hlist_node hnode;
 334	unsigned long flags;
 335	int lease_shared_gen;
 336	u32 lease_gen;
 337	u32 lease_seq;
 338	unsigned long lease_renew_after, lease_renew_from;
 339	unsigned long time;
 340	u64 offset;
 341};
 342
 343#define CEPH_DENTRY_REFERENCED		(1 << 0)
 344#define CEPH_DENTRY_LEASE_LIST		(1 << 1)
 345#define CEPH_DENTRY_SHRINK_LIST		(1 << 2)
 346#define CEPH_DENTRY_PRIMARY_LINK	(1 << 3)
 347#define CEPH_DENTRY_ASYNC_UNLINK_BIT	(4)
 348#define CEPH_DENTRY_ASYNC_UNLINK	(1 << CEPH_DENTRY_ASYNC_UNLINK_BIT)
 349#define CEPH_DENTRY_ASYNC_CREATE_BIT	(5)
 350#define CEPH_DENTRY_ASYNC_CREATE	(1 << CEPH_DENTRY_ASYNC_CREATE_BIT)
 351
 352struct ceph_inode_xattrs_info {
 353	/*
 354	 * (still encoded) xattr blob. we avoid the overhead of parsing
 355	 * this until someone actually calls getxattr, etc.
 356	 *
 357	 * blob->vec.iov_len == 4 implies there are no xattrs; blob ==
 358	 * NULL means we don't know.
 359	*/
 360	struct ceph_buffer *blob, *prealloc_blob;
 361
 362	struct rb_root index;
 363	bool dirty;
 364	int count;
 365	int names_size;
 366	int vals_size;
 367	u64 version, index_version;
 368};
 369
 370/*
 371 * Ceph inode.
 372 */
 373struct ceph_inode_info {
 374	struct netfs_inode netfs; /* Netfslib context and vfs inode */
 375	struct ceph_vino i_vino;   /* ceph ino + snap */
 376
 377	spinlock_t i_ceph_lock;
 378
 379	u64 i_version;
 380	u64 i_inline_version;
 381	u32 i_time_warp_seq;
 382
 383	unsigned long i_ceph_flags;
 384	atomic64_t i_release_count;
 385	atomic64_t i_ordered_count;
 386	atomic64_t i_complete_seq[2];
 387
 388	struct ceph_dir_layout i_dir_layout;
 389	struct ceph_file_layout i_layout;
 390	struct ceph_file_layout i_cached_layout;	// for async creates
 391	char *i_symlink;
 392
 393	/* for dirs */
 394	struct timespec64 i_rctime;
 395	u64 i_rbytes, i_rfiles, i_rsubdirs, i_rsnaps;
 396	u64 i_files, i_subdirs;
 397
 398	/* quotas */
 399	u64 i_max_bytes, i_max_files;
 400
 401	s32 i_dir_pin;
 402
 403	struct rb_root i_fragtree;
 404	int i_fragtree_nsplits;
 405	struct mutex i_fragtree_mutex;
 406
 407	struct ceph_inode_xattrs_info i_xattrs;
 408
 409	/* capabilities.  protected _both_ by i_ceph_lock and cap->session's
 410	 * s_mutex. */
 411	struct rb_root i_caps;           /* cap list */
 412	struct ceph_cap *i_auth_cap;     /* authoritative cap, if any */
 413	unsigned i_dirty_caps, i_flushing_caps;     /* mask of dirtied fields */
 414
 415	/*
 416	 * Link to the auth cap's session's s_cap_dirty list. s_cap_dirty
 417	 * is protected by the mdsc->cap_dirty_lock, but each individual item
 418	 * is also protected by the inode's i_ceph_lock. Walking s_cap_dirty
 419	 * requires the mdsc->cap_dirty_lock. List presence for an item can
 420	 * be tested under the i_ceph_lock. Changing anything requires both.
 421	 */
 422	struct list_head i_dirty_item;
 423
 424	/*
 425	 * Link to session's s_cap_flushing list. Protected in a similar
 426	 * fashion to i_dirty_item, but also by the s_mutex for changes. The
 427	 * s_cap_flushing list can be walked while holding either the s_mutex
 428	 * or msdc->cap_dirty_lock. List presence can also be checked while
 429	 * holding the i_ceph_lock for this inode.
 430	 */
 431	struct list_head i_flushing_item;
 432
 433	/* we need to track cap writeback on a per-cap-bit basis, to allow
 434	 * overlapping, pipelined cap flushes to the mds.  we can probably
 435	 * reduce the tid to 8 bits if we're concerned about inode size. */
 436	struct ceph_cap_flush *i_prealloc_cap_flush;
 437	struct list_head i_cap_flush_list;
 438	wait_queue_head_t i_cap_wq;      /* threads waiting on a capability */
 439	unsigned long i_hold_caps_max; /* jiffies */
 440	struct list_head i_cap_delay_list;  /* for delayed cap release to mds */
 441	struct ceph_cap_reservation i_cap_migration_resv;
 442	struct list_head i_cap_snaps;   /* snapped state pending flush to mds */
 443	struct ceph_snap_context *i_head_snapc;  /* set if wr_buffer_head > 0 or
 444						    dirty|flushing caps */
 445	unsigned i_snap_caps;           /* cap bits for snapped files */
 446
 447	unsigned long i_last_rd;
 448	unsigned long i_last_wr;
 449	int i_nr_by_mode[CEPH_FILE_MODE_BITS];  /* open file counts */
 450
 451	struct mutex i_truncate_mutex;
 452	u32 i_truncate_seq;        /* last truncate to smaller size */
 453	u64 i_truncate_size;       /*  and the size we last truncated down to */
 454	int i_truncate_pending;    /*  still need to call vmtruncate */
 455	/*
 456	 * For none fscrypt case it equals to i_truncate_size or it will
 457	 * equals to fscrypt_file_size
 458	 */
 459	u64 i_truncate_pagecache_size;
 460
 461	u64 i_max_size;            /* max file size authorized by mds */
 462	u64 i_reported_size; /* (max_)size reported to or requested of mds */
 463	u64 i_wanted_max_size;     /* offset we'd like to write too */
 464	u64 i_requested_max_size;  /* max_size we've requested */
 465
 466	/* held references to caps */
 467	int i_pin_ref;
 468	int i_rd_ref, i_rdcache_ref, i_wr_ref, i_wb_ref, i_fx_ref;
 469	int i_wrbuffer_ref, i_wrbuffer_ref_head;
 470	atomic_t i_filelock_ref;
 471	atomic_t i_shared_gen;       /* increment each time we get FILE_SHARED */
 472	u32 i_rdcache_gen;      /* incremented each time we get FILE_CACHE. */
 473	u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */
 474
 475	struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */
 476	struct list_head i_unsafe_iops;   /* uncommitted mds inode ops */
 477	spinlock_t i_unsafe_lock;
 478
 479	union {
 480		struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */
 481		struct ceph_snapid_map *i_snapid_map; /* snapid -> dev_t */
 482	};
 483	struct list_head i_snap_realm_item;
 484	struct list_head i_snap_flush_item;
 485	struct timespec64 i_btime;
 486	struct timespec64 i_snap_btime;
 487
 488	struct work_struct i_work;
 489	unsigned long  i_work_mask;
 490
 491#ifdef CONFIG_FS_ENCRYPTION
 492	struct fscrypt_inode_info *i_crypt_info;
 493	u32 fscrypt_auth_len;
 494	u32 fscrypt_file_len;
 495	u8 *fscrypt_auth;
 496	u8 *fscrypt_file;
 497#endif
 498};
 499
 500struct ceph_netfs_request_data {
 501	int caps;
 502
 503	/*
 504	 * Maximum size of a file readahead request.
 505	 * The fadvise could update the bdi's default ra_pages.
 506	 */
 507	unsigned int file_ra_pages;
 508
 509	/* Set it if fadvise disables file readahead entirely */
 510	bool file_ra_disabled;
 511};
 512
 513static inline struct ceph_inode_info *
 514ceph_inode(const struct inode *inode)
 515{
 516	return container_of(inode, struct ceph_inode_info, netfs.inode);
 517}
 518
 519static inline struct ceph_fs_client *
 520ceph_inode_to_fs_client(const struct inode *inode)
 521{
 522	return (struct ceph_fs_client *)inode->i_sb->s_fs_info;
 523}
 524
 525static inline struct ceph_fs_client *
 526ceph_sb_to_fs_client(const struct super_block *sb)
 527{
 528	return (struct ceph_fs_client *)sb->s_fs_info;
 529}
 530
 531static inline struct ceph_mds_client *
 532ceph_sb_to_mdsc(const struct super_block *sb)
 533{
 534	return (struct ceph_mds_client *)ceph_sb_to_fs_client(sb)->mdsc;
 535}
 536
 537static inline struct ceph_client *
 538ceph_inode_to_client(const struct inode *inode)
 539{
 540	return (struct ceph_client *)ceph_inode_to_fs_client(inode)->client;
 541}
 542
 543static inline struct ceph_vino
 544ceph_vino(const struct inode *inode)
 545{
 546	return ceph_inode(inode)->i_vino;
 547}
 548
 549static inline u32 ceph_ino_to_ino32(u64 vino)
 550{
 551	u32 ino = vino & 0xffffffff;
 552	ino ^= vino >> 32;
 553	if (!ino)
 554		ino = 2;
 555	return ino;
 556}
 557
 558/*
 559 * Inode numbers in cephfs are 64 bits, but inode->i_ino is 32-bits on
 560 * some arches. We generally do not use this value inside the ceph driver, but
 561 * we do want to set it to something, so that generic vfs code has an
 562 * appropriate value for tracepoints and the like.
 563 */
 564static inline ino_t ceph_vino_to_ino_t(struct ceph_vino vino)
 565{
 566	if (sizeof(ino_t) == sizeof(u32))
 567		return ceph_ino_to_ino32(vino.ino);
 568	return (ino_t)vino.ino;
 569}
 570
 571/* for printf-style formatting */
 572#define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap
 573
 574static inline u64 ceph_ino(struct inode *inode)
 575{
 576	return ceph_inode(inode)->i_vino.ino;
 577}
 578
 579static inline u64 ceph_snap(struct inode *inode)
 580{
 581	return ceph_inode(inode)->i_vino.snap;
 582}
 583
 584/**
 585 * ceph_present_ino - format an inode number for presentation to userland
 586 * @sb: superblock where the inode lives
 587 * @ino: inode number to (possibly) convert
 588 *
 589 * If the user mounted with the ino32 option, then the 64-bit value needs
 590 * to be converted to something that can fit inside 32 bits. Note that
 591 * internal kernel code never uses this value, so this is entirely for
 592 * userland consumption.
 593 */
 594static inline u64 ceph_present_ino(struct super_block *sb, u64 ino)
 595{
 596	if (unlikely(ceph_test_mount_opt(ceph_sb_to_fs_client(sb), INO32)))
 597		return ceph_ino_to_ino32(ino);
 598	return ino;
 599}
 600
 601static inline u64 ceph_present_inode(struct inode *inode)
 602{
 603	return ceph_present_ino(inode->i_sb, ceph_ino(inode));
 604}
 605
 606static inline int ceph_ino_compare(struct inode *inode, void *data)
 607{
 608	struct ceph_vino *pvino = (struct ceph_vino *)data;
 609	struct ceph_inode_info *ci = ceph_inode(inode);
 610	return ci->i_vino.ino == pvino->ino &&
 611		ci->i_vino.snap == pvino->snap;
 612}
 613
 614/*
 615 * The MDS reserves a set of inodes for its own usage. These should never
 616 * be accessible by clients, and so the MDS has no reason to ever hand these
 617 * out. The range is CEPH_MDS_INO_MDSDIR_OFFSET..CEPH_INO_SYSTEM_BASE.
 618 *
 619 * These come from src/mds/mdstypes.h in the ceph sources.
 620 */
 621#define CEPH_MAX_MDS			0x100
 622#define CEPH_NUM_STRAY			10
 623#define CEPH_MDS_INO_MDSDIR_OFFSET	(1 * CEPH_MAX_MDS)
 624#define CEPH_MDS_INO_LOG_OFFSET		(2 * CEPH_MAX_MDS)
 625#define CEPH_INO_SYSTEM_BASE		((6*CEPH_MAX_MDS) + (CEPH_MAX_MDS * CEPH_NUM_STRAY))
 626
 627static inline bool ceph_vino_is_reserved(const struct ceph_vino vino)
 628{
 629	if (vino.ino >= CEPH_INO_SYSTEM_BASE ||
 630	    vino.ino < CEPH_MDS_INO_MDSDIR_OFFSET)
 631		return false;
 632
 633	/* Don't warn on mdsdirs */
 634	WARN_RATELIMIT(vino.ino >= CEPH_MDS_INO_LOG_OFFSET,
 635			"Attempt to access reserved inode number 0x%llx",
 636			vino.ino);
 637	return true;
 638}
 639
 640static inline struct inode *ceph_find_inode(struct super_block *sb,
 641					    struct ceph_vino vino)
 642{
 643	if (ceph_vino_is_reserved(vino))
 644		return NULL;
 645
 646	/*
 647	 * NB: The hashval will be run through the fs/inode.c hash function
 648	 * anyway, so there is no need to squash the inode number down to
 649	 * 32-bits first. Just use low-order bits on arches with 32-bit long.
 650	 */
 651	return ilookup5(sb, (unsigned long)vino.ino, ceph_ino_compare, &vino);
 652}
 653
 654
 655/*
 656 * Ceph inode.
 657 */
 658#define CEPH_I_DIR_ORDERED	(1 << 0)  /* dentries in dir are ordered */
 659#define CEPH_I_FLUSH		(1 << 2)  /* do not delay flush of dirty metadata */
 660#define CEPH_I_POOL_PERM	(1 << 3)  /* pool rd/wr bits are valid */
 661#define CEPH_I_POOL_RD		(1 << 4)  /* can read from pool */
 662#define CEPH_I_POOL_WR		(1 << 5)  /* can write to pool */
 663#define CEPH_I_SEC_INITED	(1 << 6)  /* security initialized */
 664#define CEPH_I_KICK_FLUSH	(1 << 7)  /* kick flushing caps */
 665#define CEPH_I_FLUSH_SNAPS	(1 << 8)  /* need flush snapss */
 666#define CEPH_I_ERROR_WRITE	(1 << 9) /* have seen write errors */
 667#define CEPH_I_ERROR_FILELOCK	(1 << 10) /* have seen file lock errors */
 668#define CEPH_I_ODIRECT_BIT	(11) /* inode in direct I/O mode */
 669#define CEPH_I_ODIRECT		(1 << CEPH_I_ODIRECT_BIT)
 670#define CEPH_ASYNC_CREATE_BIT	(12)	  /* async create in flight for this */
 671#define CEPH_I_ASYNC_CREATE	(1 << CEPH_ASYNC_CREATE_BIT)
 672#define CEPH_I_SHUTDOWN		(1 << 13) /* inode is no longer usable */
 673#define CEPH_I_ASYNC_CHECK_CAPS	(1 << 14) /* check caps immediately after async
 674					     creating finishes */
 675
 676/*
 677 * Masks of ceph inode work.
 678 */
 679#define CEPH_I_WORK_WRITEBACK		0
 680#define CEPH_I_WORK_INVALIDATE_PAGES	1
 681#define CEPH_I_WORK_VMTRUNCATE		2
 682#define CEPH_I_WORK_CHECK_CAPS		3
 683#define CEPH_I_WORK_FLUSH_SNAPS		4
 684
 685/*
 686 * We set the ERROR_WRITE bit when we start seeing write errors on an inode
 687 * and then clear it when they start succeeding. Note that we do a lockless
 688 * check first, and only take the lock if it looks like it needs to be changed.
 689 * The write submission code just takes this as a hint, so we're not too
 690 * worried if a few slip through in either direction.
 691 */
 692static inline void ceph_set_error_write(struct ceph_inode_info *ci)
 693{
 694	if (!(READ_ONCE(ci->i_ceph_flags) & CEPH_I_ERROR_WRITE)) {
 695		spin_lock(&ci->i_ceph_lock);
 696		ci->i_ceph_flags |= CEPH_I_ERROR_WRITE;
 697		spin_unlock(&ci->i_ceph_lock);
 698	}
 699}
 700
 701static inline void ceph_clear_error_write(struct ceph_inode_info *ci)
 702{
 703	if (READ_ONCE(ci->i_ceph_flags) & CEPH_I_ERROR_WRITE) {
 704		spin_lock(&ci->i_ceph_lock);
 705		ci->i_ceph_flags &= ~CEPH_I_ERROR_WRITE;
 706		spin_unlock(&ci->i_ceph_lock);
 707	}
 708}
 709
 710static inline void __ceph_dir_set_complete(struct ceph_inode_info *ci,
 711					   long long release_count,
 712					   long long ordered_count)
 713{
 714	/*
 715	 * Makes sure operations that setup readdir cache (update page
 716	 * cache and i_size) are strongly ordered w.r.t. the following
 717	 * atomic64_set() operations.
 718	 */
 719	smp_mb();
 720	atomic64_set(&ci->i_complete_seq[0], release_count);
 721	atomic64_set(&ci->i_complete_seq[1], ordered_count);
 722}
 723
 724static inline void __ceph_dir_clear_complete(struct ceph_inode_info *ci)
 725{
 726	atomic64_inc(&ci->i_release_count);
 727}
 728
 729static inline void __ceph_dir_clear_ordered(struct ceph_inode_info *ci)
 730{
 731	atomic64_inc(&ci->i_ordered_count);
 732}
 733
 734static inline bool __ceph_dir_is_complete(struct ceph_inode_info *ci)
 735{
 736	return atomic64_read(&ci->i_complete_seq[0]) ==
 737		atomic64_read(&ci->i_release_count);
 738}
 739
 740static inline bool __ceph_dir_is_complete_ordered(struct ceph_inode_info *ci)
 741{
 742	return  atomic64_read(&ci->i_complete_seq[0]) ==
 743		atomic64_read(&ci->i_release_count) &&
 744		atomic64_read(&ci->i_complete_seq[1]) ==
 745		atomic64_read(&ci->i_ordered_count);
 746}
 747
 748static inline void ceph_dir_clear_complete(struct inode *inode)
 749{
 750	__ceph_dir_clear_complete(ceph_inode(inode));
 751}
 752
 753static inline void ceph_dir_clear_ordered(struct inode *inode)
 754{
 755	__ceph_dir_clear_ordered(ceph_inode(inode));
 756}
 757
 758static inline bool ceph_dir_is_complete_ordered(struct inode *inode)
 759{
 760	bool ret = __ceph_dir_is_complete_ordered(ceph_inode(inode));
 761	smp_rmb();
 762	return ret;
 763}
 764
 765/* find a specific frag @f */
 766extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci,
 767						u32 f);
 768
 769/*
 770 * choose fragment for value @v.  copy frag content to pfrag, if leaf
 771 * exists
 772 */
 773extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v,
 774			    struct ceph_inode_frag *pfrag,
 775			    int *found);
 776
 777static inline struct ceph_dentry_info *ceph_dentry(const struct dentry *dentry)
 778{
 779	return (struct ceph_dentry_info *)dentry->d_fsdata;
 780}
 781
 782/*
 783 * caps helpers
 784 */
 785static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci)
 786{
 787	return !RB_EMPTY_ROOT(&ci->i_caps);
 788}
 789
 790extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented);
 791extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t);
 792extern int __ceph_caps_issued_mask_metric(struct ceph_inode_info *ci, int mask,
 793					  int t);
 794extern int __ceph_caps_issued_other(struct ceph_inode_info *ci,
 795				    struct ceph_cap *cap);
 796
 797static inline int ceph_caps_issued(struct ceph_inode_info *ci)
 798{
 799	int issued;
 800	spin_lock(&ci->i_ceph_lock);
 801	issued = __ceph_caps_issued(ci, NULL);
 802	spin_unlock(&ci->i_ceph_lock);
 803	return issued;
 804}
 805
 806static inline int ceph_caps_issued_mask_metric(struct ceph_inode_info *ci,
 807					       int mask, int touch)
 808{
 809	int r;
 810	spin_lock(&ci->i_ceph_lock);
 811	r = __ceph_caps_issued_mask_metric(ci, mask, touch);
 812	spin_unlock(&ci->i_ceph_lock);
 813	return r;
 814}
 815
 816static inline int __ceph_caps_dirty(struct ceph_inode_info *ci)
 817{
 818	return ci->i_dirty_caps | ci->i_flushing_caps;
 819}
 820extern struct ceph_cap_flush *ceph_alloc_cap_flush(void);
 821extern void ceph_free_cap_flush(struct ceph_cap_flush *cf);
 822extern int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask,
 823				  struct ceph_cap_flush **pcf);
 824
 825extern int __ceph_caps_revoking_other(struct ceph_inode_info *ci,
 826				      struct ceph_cap *ocap, int mask);
 827extern int __ceph_caps_used(struct ceph_inode_info *ci);
 828
 829static inline bool __ceph_is_file_opened(struct ceph_inode_info *ci)
 830{
 831	return ci->i_nr_by_mode[0];
 832}
 833extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci);
 834extern int __ceph_caps_wanted(struct ceph_inode_info *ci);
 835
 836/* what the mds thinks we want */
 837extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci, bool check);
 838
 839extern void ceph_caps_init(struct ceph_mds_client *mdsc);
 840extern void ceph_caps_finalize(struct ceph_mds_client *mdsc);
 841extern void ceph_adjust_caps_max_min(struct ceph_mds_client *mdsc,
 842				     struct ceph_mount_options *fsopt);
 843extern int ceph_reserve_caps(struct ceph_mds_client *mdsc,
 844			     struct ceph_cap_reservation *ctx, int need);
 845extern void ceph_unreserve_caps(struct ceph_mds_client *mdsc,
 846			       struct ceph_cap_reservation *ctx);
 847extern void ceph_reservation_status(struct ceph_fs_client *client,
 848				    int *total, int *avail, int *used,
 849				    int *reserved, int *min);
 850extern void change_auth_cap_ses(struct ceph_inode_info *ci,
 851				struct ceph_mds_session *session);
 852
 853
 854
 855/*
 856 * we keep buffered readdir results attached to file->private_data
 857 */
 858#define CEPH_F_SYNC     1
 859#define CEPH_F_ATEND    2
 860
 861struct ceph_file_info {
 862	short fmode;     /* initialized on open */
 863	short flags;     /* CEPH_F_* */
 864
 865	spinlock_t rw_contexts_lock;
 866	struct list_head rw_contexts;
 867
 868	u32 filp_gen;
 869};
 870
 871struct ceph_dir_file_info {
 872	struct ceph_file_info file_info;
 873
 874	/* readdir: position within the dir */
 875	u32 frag;
 876	struct ceph_mds_request *last_readdir;
 877
 878	/* readdir: position within a frag */
 879	unsigned next_offset;  /* offset of next chunk (last_name's + 1) */
 880	char *last_name;       /* last entry in previous chunk */
 881	long long dir_release_count;
 882	long long dir_ordered_count;
 883	int readdir_cache_idx;
 884
 885	/* used for -o dirstat read() on directory thing */
 886	char *dir_info;
 887	int dir_info_len;
 888};
 889
 890struct ceph_rw_context {
 891	struct list_head list;
 892	struct task_struct *thread;
 893	int caps;
 894};
 895
 896#define CEPH_DEFINE_RW_CONTEXT(_name, _caps)	\
 897	struct ceph_rw_context _name = {	\
 898		.thread = current,		\
 899		.caps = _caps,			\
 900	}
 901
 902static inline void ceph_add_rw_context(struct ceph_file_info *cf,
 903				       struct ceph_rw_context *ctx)
 904{
 905	spin_lock(&cf->rw_contexts_lock);
 906	list_add(&ctx->list, &cf->rw_contexts);
 907	spin_unlock(&cf->rw_contexts_lock);
 908}
 909
 910static inline void ceph_del_rw_context(struct ceph_file_info *cf,
 911				       struct ceph_rw_context *ctx)
 912{
 913	spin_lock(&cf->rw_contexts_lock);
 914	list_del(&ctx->list);
 915	spin_unlock(&cf->rw_contexts_lock);
 916}
 917
 918static inline struct ceph_rw_context*
 919ceph_find_rw_context(struct ceph_file_info *cf)
 920{
 921	struct ceph_rw_context *ctx, *found = NULL;
 922	spin_lock(&cf->rw_contexts_lock);
 923	list_for_each_entry(ctx, &cf->rw_contexts, list) {
 924		if (ctx->thread == current) {
 925			found = ctx;
 926			break;
 927		}
 928	}
 929	spin_unlock(&cf->rw_contexts_lock);
 930	return found;
 931}
 932
 933struct ceph_readdir_cache_control {
 934	struct folio *folio;
 935	struct dentry **dentries;
 936	int index;
 937};
 938
 939/*
 940 * A "snap realm" describes a subset of the file hierarchy sharing
 941 * the same set of snapshots that apply to it.  The realms themselves
 942 * are organized into a hierarchy, such that children inherit (some of)
 943 * the snapshots of their parents.
 944 *
 945 * All inodes within the realm that have capabilities are linked into a
 946 * per-realm list.
 947 */
 948struct ceph_snap_realm {
 949	u64 ino;
 950	struct inode *inode;
 951	atomic_t nref;
 952	struct rb_node node;
 953
 954	u64 created, seq;
 955	u64 parent_ino;
 956	u64 parent_since;   /* snapid when our current parent became so */
 957
 958	u64 *prior_parent_snaps;      /* snaps inherited from any parents we */
 959	u32 num_prior_parent_snaps;   /*  had prior to parent_since */
 960	u64 *snaps;                   /* snaps specific to this realm */
 961	u32 num_snaps;
 962
 963	struct ceph_snap_realm *parent;
 964	struct list_head children;       /* list of child realms */
 965	struct list_head child_item;
 966
 967	struct list_head empty_item;     /* if i have ref==0 */
 968
 969	struct list_head dirty_item;     /* if realm needs new context */
 970
 971	struct list_head rebuild_item;   /* rebuild snap realms _downward_ in hierarchy */
 972
 973	/* the current set of snaps for this realm */
 974	struct ceph_snap_context *cached_context;
 975
 976	struct list_head inodes_with_caps;
 977	spinlock_t inodes_with_caps_lock;
 978};
 979
 980static inline int default_congestion_kb(void)
 981{
 982	int congestion_kb;
 983
 984	/*
 985	 * Copied from NFS
 986	 *
 987	 * congestion size, scale with available memory.
 988	 *
 989	 *  64MB:    8192k
 990	 * 128MB:   11585k
 991	 * 256MB:   16384k
 992	 * 512MB:   23170k
 993	 *   1GB:   32768k
 994	 *   2GB:   46340k
 995	 *   4GB:   65536k
 996	 *   8GB:   92681k
 997	 *  16GB:  131072k
 998	 *
 999	 * This allows larger machines to have larger/more transfers.
1000	 * Limit the default to 256M
1001	 */
1002	congestion_kb = (16*int_sqrt(totalram_pages())) << (PAGE_SHIFT-10);
1003	if (congestion_kb > 256*1024)
1004		congestion_kb = 256*1024;
1005
1006	return congestion_kb;
1007}
1008
1009
1010/* super.c */
1011extern int ceph_force_reconnect(struct super_block *sb);
1012/* snap.c */
1013struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc,
1014					       u64 ino);
1015extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc,
1016				struct ceph_snap_realm *realm);
1017extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc,
1018				struct ceph_snap_realm *realm);
1019extern int ceph_update_snap_trace(struct ceph_mds_client *m,
1020				  void *p, void *e, bool deletion,
1021				  struct ceph_snap_realm **realm_ret);
1022void ceph_change_snap_realm(struct inode *inode, struct ceph_snap_realm *realm);
1023extern void ceph_handle_snap(struct ceph_mds_client *mdsc,
1024			     struct ceph_mds_session *session,
1025			     struct ceph_msg *msg);
1026extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci,
1027				  struct ceph_cap_snap *capsnap);
1028extern void ceph_cleanup_global_and_empty_realms(struct ceph_mds_client *mdsc);
1029
1030extern struct ceph_snapid_map *ceph_get_snapid_map(struct ceph_mds_client *mdsc,
1031						   u64 snap);
1032extern void ceph_put_snapid_map(struct ceph_mds_client* mdsc,
1033				struct ceph_snapid_map *sm);
1034extern void ceph_trim_snapid_map(struct ceph_mds_client *mdsc);
1035extern void ceph_cleanup_snapid_map(struct ceph_mds_client *mdsc);
1036void ceph_umount_begin(struct super_block *sb);
1037
1038
1039/*
1040 * a cap_snap is "pending" if it is still awaiting an in-progress
1041 * sync write (that may/may not still update size, mtime, etc.).
1042 */
1043static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci)
1044{
1045	return !list_empty(&ci->i_cap_snaps) &&
1046	       list_last_entry(&ci->i_cap_snaps, struct ceph_cap_snap,
1047			       ci_item)->writing;
1048}
1049
1050/* inode.c */
1051struct ceph_mds_reply_info_in;
1052struct ceph_mds_reply_dirfrag;
1053struct ceph_acl_sec_ctx;
1054
1055extern const struct inode_operations ceph_file_iops;
1056
1057extern struct inode *ceph_alloc_inode(struct super_block *sb);
1058extern void ceph_evict_inode(struct inode *inode);
1059extern void ceph_free_inode(struct inode *inode);
1060
1061struct inode *ceph_new_inode(struct inode *dir, struct dentry *dentry,
1062			     umode_t *mode, struct ceph_acl_sec_ctx *as_ctx);
1063void ceph_as_ctx_to_req(struct ceph_mds_request *req,
1064			struct ceph_acl_sec_ctx *as_ctx);
1065
1066extern struct inode *ceph_get_inode(struct super_block *sb,
1067				    struct ceph_vino vino,
1068				    struct inode *newino);
1069extern struct inode *ceph_get_snapdir(struct inode *parent);
1070extern int ceph_fill_file_size(struct inode *inode, int issued,
1071			       u32 truncate_seq, u64 truncate_size, u64 size);
1072extern void ceph_fill_file_time(struct inode *inode, int issued,
1073				u64 time_warp_seq, struct timespec64 *ctime,
1074				struct timespec64 *mtime,
1075				struct timespec64 *atime);
1076extern int ceph_fill_inode(struct inode *inode, struct page *locked_page,
1077		    struct ceph_mds_reply_info_in *iinfo,
1078		    struct ceph_mds_reply_dirfrag *dirinfo,
1079		    struct ceph_mds_session *session, int cap_fmode,
1080		    struct ceph_cap_reservation *caps_reservation);
1081extern int ceph_fill_trace(struct super_block *sb,
1082			   struct ceph_mds_request *req);
1083extern int ceph_readdir_prepopulate(struct ceph_mds_request *req,
1084				    struct ceph_mds_session *session);
1085
1086extern bool ceph_inode_set_size(struct inode *inode, loff_t size);
1087extern void __ceph_do_pending_vmtruncate(struct inode *inode);
1088
1089void ceph_queue_inode_work(struct inode *inode, int work_bit);
1090
1091static inline void ceph_queue_vmtruncate(struct inode *inode)
1092{
1093	ceph_queue_inode_work(inode, CEPH_I_WORK_VMTRUNCATE);
1094}
1095
1096static inline void ceph_queue_invalidate(struct inode *inode)
1097{
1098	ceph_queue_inode_work(inode, CEPH_I_WORK_INVALIDATE_PAGES);
1099}
1100
1101static inline void ceph_queue_writeback(struct inode *inode)
1102{
1103	ceph_queue_inode_work(inode, CEPH_I_WORK_WRITEBACK);
1104}
1105
1106static inline void ceph_queue_check_caps(struct inode *inode)
1107{
1108	ceph_queue_inode_work(inode, CEPH_I_WORK_CHECK_CAPS);
1109}
1110
1111static inline void ceph_queue_flush_snaps(struct inode *inode)
1112{
1113	ceph_queue_inode_work(inode, CEPH_I_WORK_FLUSH_SNAPS);
1114}
1115
1116extern int ceph_try_to_choose_auth_mds(struct inode *inode, int mask);
1117extern int __ceph_do_getattr(struct inode *inode, struct page *locked_page,
1118			     int mask, bool force);
1119static inline int ceph_do_getattr(struct inode *inode, int mask, bool force)
1120{
1121	return __ceph_do_getattr(inode, NULL, mask, force);
1122}
1123extern int ceph_permission(struct mnt_idmap *idmap,
1124			   struct inode *inode, int mask);
1125
1126struct ceph_iattr {
1127	struct ceph_fscrypt_auth	*fscrypt_auth;
1128};
1129
1130extern int __ceph_setattr(struct mnt_idmap *idmap, struct inode *inode,
1131			  struct iattr *attr, struct ceph_iattr *cia);
1132extern int ceph_setattr(struct mnt_idmap *idmap,
1133			struct dentry *dentry, struct iattr *attr);
1134extern int ceph_getattr(struct mnt_idmap *idmap,
1135			const struct path *path, struct kstat *stat,
1136			u32 request_mask, unsigned int flags);
1137void ceph_inode_shutdown(struct inode *inode);
1138
1139static inline bool ceph_inode_is_shutdown(struct inode *inode)
1140{
1141	unsigned long flags = READ_ONCE(ceph_inode(inode)->i_ceph_flags);
1142	struct ceph_fs_client *fsc = ceph_inode_to_fs_client(inode);
1143	int state = READ_ONCE(fsc->mount_state);
1144
1145	return (flags & CEPH_I_SHUTDOWN) || state >= CEPH_MOUNT_SHUTDOWN;
1146}
1147
1148/* xattr.c */
1149int __ceph_setxattr(struct inode *, const char *, const void *, size_t, int);
1150int ceph_do_getvxattr(struct inode *inode, const char *name, void *value, size_t size);
1151ssize_t __ceph_getxattr(struct inode *, const char *, void *, size_t);
1152extern ssize_t ceph_listxattr(struct dentry *, char *, size_t);
1153extern struct ceph_buffer *__ceph_build_xattrs_blob(struct ceph_inode_info *ci);
1154extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci);
1155extern const struct xattr_handler * const ceph_xattr_handlers[];
1156
1157struct ceph_acl_sec_ctx {
1158#ifdef CONFIG_CEPH_FS_POSIX_ACL
1159	void *default_acl;
1160	void *acl;
1161#endif
1162#ifdef CONFIG_CEPH_FS_SECURITY_LABEL
1163	struct lsm_context lsmctx;
1164#endif
1165#ifdef CONFIG_FS_ENCRYPTION
1166	struct ceph_fscrypt_auth *fscrypt_auth;
1167#endif
1168	struct ceph_pagelist *pagelist;
1169};
1170
1171#ifdef CONFIG_SECURITY
1172extern bool ceph_security_xattr_deadlock(struct inode *in);
1173extern bool ceph_security_xattr_wanted(struct inode *in);
1174#else
1175static inline bool ceph_security_xattr_deadlock(struct inode *in)
1176{
1177	return false;
1178}
1179static inline bool ceph_security_xattr_wanted(struct inode *in)
1180{
1181	return false;
1182}
1183#endif
1184
1185#ifdef CONFIG_CEPH_FS_SECURITY_LABEL
1186extern int ceph_security_init_secctx(struct dentry *dentry, umode_t mode,
1187				     struct ceph_acl_sec_ctx *ctx);
1188static inline void ceph_security_invalidate_secctx(struct inode *inode)
1189{
1190	security_inode_invalidate_secctx(inode);
1191}
1192#else
1193static inline int ceph_security_init_secctx(struct dentry *dentry, umode_t mode,
1194					    struct ceph_acl_sec_ctx *ctx)
1195{
1196	return 0;
1197}
1198static inline void ceph_security_invalidate_secctx(struct inode *inode)
1199{
1200}
1201#endif
1202
1203void ceph_release_acl_sec_ctx(struct ceph_acl_sec_ctx *as_ctx);
1204
1205/* acl.c */
1206#ifdef CONFIG_CEPH_FS_POSIX_ACL
1207
1208struct posix_acl *ceph_get_acl(struct inode *, int, bool);
1209int ceph_set_acl(struct mnt_idmap *idmap,
1210		 struct dentry *dentry, struct posix_acl *acl, int type);
1211int ceph_pre_init_acls(struct inode *dir, umode_t *mode,
1212		       struct ceph_acl_sec_ctx *as_ctx);
1213void ceph_init_inode_acls(struct inode *inode,
1214			  struct ceph_acl_sec_ctx *as_ctx);
1215
1216static inline void ceph_forget_all_cached_acls(struct inode *inode)
1217{
1218       forget_all_cached_acls(inode);
1219}
1220
1221#else
1222
1223#define ceph_get_acl NULL
1224#define ceph_set_acl NULL
1225
1226static inline int ceph_pre_init_acls(struct inode *dir, umode_t *mode,
1227				     struct ceph_acl_sec_ctx *as_ctx)
1228{
1229	return 0;
1230}
1231static inline void ceph_init_inode_acls(struct inode *inode,
1232					struct ceph_acl_sec_ctx *as_ctx)
1233{
1234}
1235
1236static inline void ceph_forget_all_cached_acls(struct inode *inode)
1237{
1238}
1239
1240#endif
1241
1242/* caps.c */
1243extern const char *ceph_cap_string(int c);
1244extern void ceph_handle_caps(struct ceph_mds_session *session,
1245			     struct ceph_msg *msg);
1246extern struct ceph_cap *ceph_get_cap(struct ceph_mds_client *mdsc,
1247				     struct ceph_cap_reservation *ctx);
1248extern void ceph_add_cap(struct inode *inode,
1249			 struct ceph_mds_session *session, u64 cap_id,
1250			 unsigned issued, unsigned wanted,
1251			 unsigned cap, unsigned seq, u64 realmino, int flags,
1252			 struct ceph_cap **new_cap);
1253extern void __ceph_remove_cap(struct ceph_cap *cap, bool queue_release);
1254extern void ceph_remove_cap(struct ceph_mds_client *mdsc, struct ceph_cap *cap,
1255			    bool queue_release);
1256extern void __ceph_remove_caps(struct ceph_inode_info *ci);
1257extern void ceph_put_cap(struct ceph_mds_client *mdsc,
1258			 struct ceph_cap *cap);
1259extern int ceph_is_any_caps(struct inode *inode);
1260
1261extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc);
1262extern int ceph_fsync(struct file *file, loff_t start, loff_t end,
1263		      int datasync);
1264extern void ceph_early_kick_flushing_caps(struct ceph_mds_client *mdsc,
1265					  struct ceph_mds_session *session);
1266extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc,
1267				    struct ceph_mds_session *session);
1268void ceph_kick_flushing_inode_caps(struct ceph_mds_session *session,
1269				   struct ceph_inode_info *ci);
1270extern struct ceph_cap *__get_cap_for_mds(struct ceph_inode_info *ci,
1271					  int mds);
1272extern struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci,
1273					     int mds);
1274extern void ceph_take_cap_refs(struct ceph_inode_info *ci, int caps,
1275				bool snap_rwsem_locked);
1276extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps);
1277extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had);
1278extern void ceph_put_cap_refs_async(struct ceph_inode_info *ci, int had);
1279extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
1280				       struct ceph_snap_context *snapc);
1281extern void __ceph_remove_capsnap(struct inode *inode,
1282				  struct ceph_cap_snap *capsnap,
1283				  bool *wake_ci, bool *wake_mdsc);
1284extern void ceph_remove_capsnap(struct inode *inode,
1285				struct ceph_cap_snap *capsnap,
1286				bool *wake_ci, bool *wake_mdsc);
1287extern void ceph_flush_snaps(struct ceph_inode_info *ci,
1288			     struct ceph_mds_session **psession);
1289extern bool __ceph_should_report_size(struct ceph_inode_info *ci);
1290extern void ceph_check_caps(struct ceph_inode_info *ci, int flags);
1291extern unsigned long ceph_check_delayed_caps(struct ceph_mds_client *mdsc);
1292extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc);
1293extern void ceph_flush_cap_releases(struct ceph_mds_client *mdsc);
1294extern int  ceph_drop_caps_for_unlink(struct inode *inode);
1295extern int ceph_encode_inode_release(void **p, struct inode *inode,
1296				     int mds, int drop, int unless, int force);
1297extern int ceph_encode_dentry_release(void **p, struct dentry *dn,
1298				      struct inode *dir,
1299				      int mds, int drop, int unless);
1300
1301extern int __ceph_get_caps(struct inode *inode, struct ceph_file_info *fi,
1302			   int need, int want, loff_t endoff, int *got);
1303extern int ceph_get_caps(struct file *filp, int need, int want,
1304			 loff_t endoff, int *got);
1305extern int ceph_try_get_caps(struct inode *inode,
1306			     int need, int want, bool nonblock, int *got);
1307
1308/* for counting open files by mode */
1309extern void ceph_get_fmode(struct ceph_inode_info *ci, int mode, int count);
1310extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode, int count);
1311extern void __ceph_touch_fmode(struct ceph_inode_info *ci,
1312			       struct ceph_mds_client *mdsc, int fmode);
1313
1314/* addr.c */
1315extern const struct address_space_operations ceph_aops;
1316extern const struct netfs_request_ops ceph_netfs_ops;
1317int ceph_mmap_prepare(struct vm_area_desc *desc);
1318extern int ceph_uninline_data(struct file *file);
1319extern int ceph_pool_perm_check(struct inode *inode, int need);
1320extern void ceph_pool_perm_destroy(struct ceph_mds_client* mdsc);
1321int ceph_purge_inode_cap(struct inode *inode, struct ceph_cap *cap, bool *invalidate);
1322
1323static inline bool ceph_has_inline_data(struct ceph_inode_info *ci)
1324{
1325	if (ci->i_inline_version == CEPH_INLINE_NONE ||
1326	    ci->i_inline_version == 1) /* initial version, no data */
1327		return false;
1328	return true;
1329}
1330
1331/* file.c */
1332extern const struct file_operations ceph_file_fops;
1333
1334extern int ceph_renew_caps(struct inode *inode, int fmode);
1335extern int ceph_open(struct inode *inode, struct file *file);
1336extern int ceph_atomic_open(struct inode *dir, struct dentry *dentry,
1337			    struct file *file, unsigned flags, umode_t mode);
1338extern ssize_t __ceph_sync_read(struct inode *inode, loff_t *ki_pos,
1339				struct iov_iter *to, int *retry_op,
1340				u64 *last_objver);
1341extern int ceph_release(struct inode *inode, struct file *filp);
1342extern void ceph_fill_inline_data(struct inode *inode, struct page *locked_page,
1343				  char *data, size_t len);
1344
1345/* dir.c */
1346extern const struct file_operations ceph_dir_fops;
1347extern const struct file_operations ceph_snapdir_fops;
1348extern const struct inode_operations ceph_dir_iops;
1349extern const struct inode_operations ceph_snapdir_iops;
1350extern const struct dentry_operations ceph_dentry_ops;
1351
1352extern loff_t ceph_make_fpos(unsigned high, unsigned off, bool hash_order);
1353extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry);
1354extern struct dentry *ceph_handle_snapdir(struct ceph_mds_request *req,
1355			       struct dentry *dentry);
1356extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req,
1357					 struct dentry *dentry, int err);
1358
1359extern void __ceph_dentry_lease_touch(struct ceph_dentry_info *di);
1360extern void __ceph_dentry_dir_lease_touch(struct ceph_dentry_info *di);
1361extern void ceph_invalidate_dentry_lease(struct dentry *dentry);
1362extern int ceph_trim_dentries(struct ceph_mds_client *mdsc);
1363extern unsigned ceph_dentry_hash(struct inode *dir, struct dentry *dn);
1364extern void ceph_readdir_cache_release(struct ceph_readdir_cache_control *ctl);
1365
1366/* ioctl.c */
1367extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
1368
1369/* export.c */
1370extern const struct export_operations ceph_export_ops;
1371struct inode *ceph_lookup_inode(struct super_block *sb, u64 ino);
1372
1373/* locks.c */
1374extern __init void ceph_flock_init(void);
1375extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl);
1376extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl);
1377extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num);
1378extern int ceph_encode_locks_to_buffer(struct inode *inode,
1379				       struct ceph_filelock *flocks,
1380				       int num_fcntl_locks,
1381				       int num_flock_locks);
1382extern int ceph_locks_to_pagelist(struct ceph_filelock *flocks,
1383				  struct ceph_pagelist *pagelist,
1384				  int num_fcntl_locks, int num_flock_locks);
1385
1386/* debugfs.c */
1387extern void ceph_fs_debugfs_init(struct ceph_fs_client *client);
1388extern void ceph_fs_debugfs_cleanup(struct ceph_fs_client *client);
1389
1390/* quota.c */
1391
1392enum quota_get_realm {
1393	QUOTA_GET_MAX_FILES,
1394	QUOTA_GET_MAX_BYTES,
1395	QUOTA_GET_ANY
1396};
1397
1398static inline bool __ceph_has_quota(struct ceph_inode_info *ci,
1399				    enum quota_get_realm which)
1400{
1401	bool has_quota = false;
1402
1403	switch (which) {
1404	case QUOTA_GET_MAX_BYTES:
1405		has_quota = !!ci->i_max_bytes;
1406		break;
1407	case QUOTA_GET_MAX_FILES:
1408		has_quota = !!ci->i_max_files;
1409		break;
1410	default:
1411		has_quota = !!(ci->i_max_files || ci->i_max_bytes);
1412	}
1413	return has_quota;
1414}
1415
1416extern void ceph_adjust_quota_realms_count(struct inode *inode, bool inc);
1417
1418static inline void __ceph_update_quota(struct ceph_inode_info *ci,
1419				       u64 max_bytes, u64 max_files)
1420{
1421	bool had_quota, has_quota;
1422	had_quota = __ceph_has_quota(ci, QUOTA_GET_ANY);
1423	ci->i_max_bytes = max_bytes;
1424	ci->i_max_files = max_files;
1425	has_quota = __ceph_has_quota(ci, QUOTA_GET_ANY);
1426
1427	if (had_quota != has_quota)
1428		ceph_adjust_quota_realms_count(&ci->netfs.inode, has_quota);
1429}
1430
1431static inline int __ceph_sparse_read_ext_count(struct inode *inode, u64 len)
1432{
1433	int cnt = 0;
1434
1435	if (IS_ENCRYPTED(inode)) {
1436		cnt = len >> CEPH_FSCRYPT_BLOCK_SHIFT;
1437		if (cnt > CEPH_SPARSE_EXT_ARRAY_INITIAL)
1438			cnt = 0;
1439	}
1440
1441	return cnt;
1442}
1443
1444extern void ceph_handle_quota(struct ceph_mds_client *mdsc,
1445			      struct ceph_mds_session *session,
1446			      struct ceph_msg *msg);
1447extern bool ceph_quota_is_max_files_exceeded(struct inode *inode);
1448extern bool ceph_quota_is_same_realm(struct inode *old, struct inode *new);
1449extern bool ceph_quota_is_max_bytes_exceeded(struct inode *inode,
1450					     loff_t newlen);
1451extern bool ceph_quota_is_max_bytes_approaching(struct inode *inode,
1452						loff_t newlen);
1453extern bool ceph_quota_update_statfs(struct ceph_fs_client *fsc,
1454				     struct kstatfs *buf);
1455extern void ceph_cleanup_quotarealms_inodes(struct ceph_mds_client *mdsc);
1456
1457bool ceph_inc_mds_stopping_blocker(struct ceph_mds_client *mdsc,
1458			       struct ceph_mds_session *session);
1459void ceph_dec_mds_stopping_blocker(struct ceph_mds_client *mdsc);
1460bool ceph_inc_osd_stopping_blocker(struct ceph_mds_client *mdsc);
1461void ceph_dec_osd_stopping_blocker(struct ceph_mds_client *mdsc);
1462#endif /* _FS_CEPH_SUPER_H */