1437 lines · c
1/* SPDX-License-Identifier: GPL-2.0 */2#ifndef _FS_CEPH_SUPER_H3#define _FS_CEPH_SUPER_H4 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 facilitate29 * 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_##opt56#define ceph_clear_mount_opt(fsc, opt) \57 (fsc)->mount_options->flags &= ~CEPH_MOUNT_OPT_##opt58#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_LEN63/* osd has a configurable limitaion of max write size.64 * CEPH_MSG_MAX_DATA_LEN should be small enough. */65#define CEPH_MAX_WRITE_SIZE CEPH_MSG_MAX_DATA_LEN66#define CEPH_RASIZE_DEFAULT (8192*1024) /* max readahead */67#define CEPH_MAX_READDIR_DEFAULT 102468#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 reopen73 * the file. Delay a minimum amount of time, even if we send a cap74 * message for some other reason. Otherwise, take the oppotunity to75 * 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 below96 * 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/* mount state */108enum {109 CEPH_MOUNT_MOUNTING,110 CEPH_MOUNT_MOUNTED,111 CEPH_MOUNT_UNMOUNTING,112 CEPH_MOUNT_UNMOUNTED,113 CEPH_MOUNT_SHUTDOWN,114 CEPH_MOUNT_RECOVER,115 CEPH_MOUNT_FENCE_IO,116};117 118#define CEPH_ASYNC_CREATE_CONFLICT_BITS 8119 120struct ceph_fs_client {121 struct super_block *sb;122 123 struct list_head metric_wakeup;124 125 struct ceph_mount_options *mount_options;126 struct ceph_client *client;127 128 int mount_state;129 130 bool blocklisted;131 132 bool have_copy_from2;133 134 u32 filp_gen;135 loff_t max_file_size;136 137 struct ceph_mds_client *mdsc;138 139 atomic_long_t writeback_count;140 bool write_congested;141 142 struct workqueue_struct *inode_wq;143 struct workqueue_struct *cap_wq;144 145 DECLARE_HASHTABLE(async_unlink_conflict, CEPH_ASYNC_CREATE_CONFLICT_BITS);146 spinlock_t async_unlink_conflict_lock;147 148#ifdef CONFIG_DEBUG_FS149 struct dentry *debugfs_dentry_lru, *debugfs_caps;150 struct dentry *debugfs_congestion_kb;151 struct dentry *debugfs_bdi;152 struct dentry *debugfs_mdsc, *debugfs_mdsmap;153 struct dentry *debugfs_status;154 struct dentry *debugfs_mds_sessions;155 struct dentry *debugfs_metrics_dir;156#endif157 158#ifdef CONFIG_CEPH_FSCACHE159 struct fscache_volume *fscache;160#endif161#ifdef CONFIG_FS_ENCRYPTION162 struct fscrypt_dummy_policy fsc_dummy_enc_policy;163#endif164};165 166/*167 * File i/o capability. This tracks shared state with the metadata168 * server that allows us to cache or writeback attributes or to read169 * and write data. For any given inode, we should have one or more170 * capabilities, one issued by each metadata server, and our171 * cumulative access is the OR of all issued capabilities.172 *173 * Each cap is referenced by the inode's i_caps rbtree and by per-mds174 * session capability lists.175 */176struct ceph_cap {177 struct ceph_inode_info *ci;178 struct rb_node ci_node; /* per-ci cap tree */179 struct ceph_mds_session *session;180 struct list_head session_caps; /* per-session caplist */181 u64 cap_id; /* unique cap id (mds provided) */182 union {183 /* in-use caps */184 struct {185 int issued; /* latest, from the mds */186 int implemented; /* implemented superset of187 issued (for revocation) */188 int mds; /* mds index for this cap */189 int mds_wanted; /* caps wanted from this mds */190 };191 /* caps to release */192 struct {193 u64 cap_ino;194 int queue_release;195 };196 };197 u32 seq, issue_seq, mseq;198 u32 cap_gen; /* active/stale cycle */199 unsigned long last_used;200 struct list_head caps_item;201};202 203#define CHECK_CAPS_AUTHONLY 1 /* only check auth cap */204#define CHECK_CAPS_FLUSH 2 /* flush any dirty caps */205#define CHECK_CAPS_NOINVAL 4 /* don't invalidate pagecache */206#define CHECK_CAPS_FLUSH_FORCE 8 /* force flush any caps */207 208struct ceph_cap_flush {209 u64 tid;210 int caps;211 bool wake; /* wake up flush waiters when finish ? */212 bool is_capsnap; /* true means capsnap */213 struct list_head g_list; // global214 struct list_head i_list; // per inode215};216 217/*218 * Snapped cap state that is pending flush to mds. When a snapshot occurs,219 * we first complete any in-process sync writes and writeback any dirty220 * data before flushing the snapped state (tracked here) back to the MDS.221 */222struct ceph_cap_snap {223 refcount_t nref;224 struct list_head ci_item;225 226 struct ceph_cap_flush cap_flush;227 228 u64 follows;229 int issued, dirty;230 struct ceph_snap_context *context;231 232 umode_t mode;233 kuid_t uid;234 kgid_t gid;235 236 struct ceph_buffer *xattr_blob;237 u64 xattr_version;238 239 u64 size;240 u64 change_attr;241 struct timespec64 mtime, atime, ctime, btime;242 u64 time_warp_seq;243 u64 truncate_size;244 u32 truncate_seq;245 int writing; /* a sync write is still in progress */246 int dirty_pages; /* dirty pages awaiting writeback */247 bool inline_data;248 bool need_flush;249};250 251static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap)252{253 if (refcount_dec_and_test(&capsnap->nref)) {254 if (capsnap->xattr_blob)255 ceph_buffer_put(capsnap->xattr_blob);256 kmem_cache_free(ceph_cap_snap_cachep, capsnap);257 }258}259 260/*261 * The frag tree describes how a directory is fragmented, potentially across262 * multiple metadata servers. It is also used to indicate points where263 * metadata authority is delegated, and whether/where metadata is replicated.264 *265 * A _leaf_ frag will be present in the i_fragtree IFF there is266 * delegation info. That is, if mds >= 0 || ndist > 0.267 */268#define CEPH_MAX_DIRFRAG_REP 4269 270struct ceph_inode_frag {271 struct rb_node node;272 273 /* fragtree state */274 u32 frag;275 int split_by; /* i.e. 2^(split_by) children */276 277 /* delegation and replication info */278 int mds; /* -1 if same authority as parent */279 int ndist; /* >0 if replicated */280 int dist[CEPH_MAX_DIRFRAG_REP];281};282 283/*284 * We cache inode xattrs as an encoded blob until they are first used,285 * at which point we parse them into an rbtree.286 */287struct ceph_inode_xattr {288 struct rb_node node;289 290 const char *name;291 int name_len;292 const char *val;293 int val_len;294 int dirty;295 296 int should_free_name;297 int should_free_val;298};299 300/*301 * Ceph dentry state302 */303struct ceph_dentry_info {304 struct dentry *dentry;305 struct ceph_mds_session *lease_session;306 struct list_head lease_list;307 struct hlist_node hnode;308 unsigned long flags;309 int lease_shared_gen;310 u32 lease_gen;311 u32 lease_seq;312 unsigned long lease_renew_after, lease_renew_from;313 unsigned long time;314 u64 offset;315};316 317#define CEPH_DENTRY_REFERENCED (1 << 0)318#define CEPH_DENTRY_LEASE_LIST (1 << 1)319#define CEPH_DENTRY_SHRINK_LIST (1 << 2)320#define CEPH_DENTRY_PRIMARY_LINK (1 << 3)321#define CEPH_DENTRY_ASYNC_UNLINK_BIT (4)322#define CEPH_DENTRY_ASYNC_UNLINK (1 << CEPH_DENTRY_ASYNC_UNLINK_BIT)323#define CEPH_DENTRY_ASYNC_CREATE_BIT (5)324#define CEPH_DENTRY_ASYNC_CREATE (1 << CEPH_DENTRY_ASYNC_CREATE_BIT)325 326struct ceph_inode_xattrs_info {327 /*328 * (still encoded) xattr blob. we avoid the overhead of parsing329 * this until someone actually calls getxattr, etc.330 *331 * blob->vec.iov_len == 4 implies there are no xattrs; blob ==332 * NULL means we don't know.333 */334 struct ceph_buffer *blob, *prealloc_blob;335 336 struct rb_root index;337 bool dirty;338 int count;339 int names_size;340 int vals_size;341 u64 version, index_version;342};343 344/*345 * Ceph inode.346 */347struct ceph_inode_info {348 struct netfs_inode netfs; /* Netfslib context and vfs inode */349 struct ceph_vino i_vino; /* ceph ino + snap */350 351 spinlock_t i_ceph_lock;352 353 u64 i_version;354 u64 i_inline_version;355 u32 i_time_warp_seq;356 357 unsigned long i_ceph_flags;358 atomic64_t i_release_count;359 atomic64_t i_ordered_count;360 atomic64_t i_complete_seq[2];361 362 struct ceph_dir_layout i_dir_layout;363 struct ceph_file_layout i_layout;364 struct ceph_file_layout i_cached_layout; // for async creates365 char *i_symlink;366 367 /* for dirs */368 struct timespec64 i_rctime;369 u64 i_rbytes, i_rfiles, i_rsubdirs, i_rsnaps;370 u64 i_files, i_subdirs;371 372 /* quotas */373 u64 i_max_bytes, i_max_files;374 375 s32 i_dir_pin;376 377 struct rb_root i_fragtree;378 int i_fragtree_nsplits;379 struct mutex i_fragtree_mutex;380 381 struct ceph_inode_xattrs_info i_xattrs;382 383 /* capabilities. protected _both_ by i_ceph_lock and cap->session's384 * s_mutex. */385 struct rb_root i_caps; /* cap list */386 struct ceph_cap *i_auth_cap; /* authoritative cap, if any */387 unsigned i_dirty_caps, i_flushing_caps; /* mask of dirtied fields */388 389 /*390 * Link to the auth cap's session's s_cap_dirty list. s_cap_dirty391 * is protected by the mdsc->cap_dirty_lock, but each individual item392 * is also protected by the inode's i_ceph_lock. Walking s_cap_dirty393 * requires the mdsc->cap_dirty_lock. List presence for an item can394 * be tested under the i_ceph_lock. Changing anything requires both.395 */396 struct list_head i_dirty_item;397 398 /*399 * Link to session's s_cap_flushing list. Protected in a similar400 * fashion to i_dirty_item, but also by the s_mutex for changes. The401 * s_cap_flushing list can be walked while holding either the s_mutex402 * or msdc->cap_dirty_lock. List presence can also be checked while403 * holding the i_ceph_lock for this inode.404 */405 struct list_head i_flushing_item;406 407 /* we need to track cap writeback on a per-cap-bit basis, to allow408 * overlapping, pipelined cap flushes to the mds. we can probably409 * reduce the tid to 8 bits if we're concerned about inode size. */410 struct ceph_cap_flush *i_prealloc_cap_flush;411 struct list_head i_cap_flush_list;412 wait_queue_head_t i_cap_wq; /* threads waiting on a capability */413 unsigned long i_hold_caps_max; /* jiffies */414 struct list_head i_cap_delay_list; /* for delayed cap release to mds */415 struct ceph_cap_reservation i_cap_migration_resv;416 struct list_head i_cap_snaps; /* snapped state pending flush to mds */417 struct ceph_snap_context *i_head_snapc; /* set if wr_buffer_head > 0 or418 dirty|flushing caps */419 unsigned i_snap_caps; /* cap bits for snapped files */420 421 unsigned long i_last_rd;422 unsigned long i_last_wr;423 int i_nr_by_mode[CEPH_FILE_MODE_BITS]; /* open file counts */424 425 struct mutex i_truncate_mutex;426 u32 i_truncate_seq; /* last truncate to smaller size */427 u64 i_truncate_size; /* and the size we last truncated down to */428 int i_truncate_pending; /* still need to call vmtruncate */429 /*430 * For none fscrypt case it equals to i_truncate_size or it will431 * equals to fscrypt_file_size432 */433 u64 i_truncate_pagecache_size;434 435 u64 i_max_size; /* max file size authorized by mds */436 u64 i_reported_size; /* (max_)size reported to or requested of mds */437 u64 i_wanted_max_size; /* offset we'd like to write too */438 u64 i_requested_max_size; /* max_size we've requested */439 440 /* held references to caps */441 int i_pin_ref;442 int i_rd_ref, i_rdcache_ref, i_wr_ref, i_wb_ref, i_fx_ref;443 int i_wrbuffer_ref, i_wrbuffer_ref_head;444 atomic_t i_filelock_ref;445 atomic_t i_shared_gen; /* increment each time we get FILE_SHARED */446 u32 i_rdcache_gen; /* incremented each time we get FILE_CACHE. */447 u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */448 449 struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */450 struct list_head i_unsafe_iops; /* uncommitted mds inode ops */451 spinlock_t i_unsafe_lock;452 453 union {454 struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */455 struct ceph_snapid_map *i_snapid_map; /* snapid -> dev_t */456 };457 struct list_head i_snap_realm_item;458 struct list_head i_snap_flush_item;459 struct timespec64 i_btime;460 struct timespec64 i_snap_btime;461 462 struct work_struct i_work;463 unsigned long i_work_mask;464 465#ifdef CONFIG_FS_ENCRYPTION466 u32 fscrypt_auth_len;467 u32 fscrypt_file_len;468 u8 *fscrypt_auth;469 u8 *fscrypt_file;470#endif471};472 473struct ceph_netfs_request_data {474 int caps;475 476 /*477 * Maximum size of a file readahead request.478 * The fadvise could update the bdi's default ra_pages.479 */480 unsigned int file_ra_pages;481 482 /* Set it if fadvise disables file readahead entirely */483 bool file_ra_disabled;484};485 486static inline struct ceph_inode_info *487ceph_inode(const struct inode *inode)488{489 return container_of(inode, struct ceph_inode_info, netfs.inode);490}491 492static inline struct ceph_fs_client *493ceph_inode_to_fs_client(const struct inode *inode)494{495 return (struct ceph_fs_client *)inode->i_sb->s_fs_info;496}497 498static inline struct ceph_fs_client *499ceph_sb_to_fs_client(const struct super_block *sb)500{501 return (struct ceph_fs_client *)sb->s_fs_info;502}503 504static inline struct ceph_mds_client *505ceph_sb_to_mdsc(const struct super_block *sb)506{507 return (struct ceph_mds_client *)ceph_sb_to_fs_client(sb)->mdsc;508}509 510static inline struct ceph_client *511ceph_inode_to_client(const struct inode *inode)512{513 return (struct ceph_client *)ceph_inode_to_fs_client(inode)->client;514}515 516static inline struct ceph_vino517ceph_vino(const struct inode *inode)518{519 return ceph_inode(inode)->i_vino;520}521 522static inline u32 ceph_ino_to_ino32(u64 vino)523{524 u32 ino = vino & 0xffffffff;525 ino ^= vino >> 32;526 if (!ino)527 ino = 2;528 return ino;529}530 531/*532 * Inode numbers in cephfs are 64 bits, but inode->i_ino is 32-bits on533 * some arches. We generally do not use this value inside the ceph driver, but534 * we do want to set it to something, so that generic vfs code has an535 * appropriate value for tracepoints and the like.536 */537static inline ino_t ceph_vino_to_ino_t(struct ceph_vino vino)538{539 if (sizeof(ino_t) == sizeof(u32))540 return ceph_ino_to_ino32(vino.ino);541 return (ino_t)vino.ino;542}543 544/* for printf-style formatting */545#define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap546 547static inline u64 ceph_ino(struct inode *inode)548{549 return ceph_inode(inode)->i_vino.ino;550}551 552static inline u64 ceph_snap(struct inode *inode)553{554 return ceph_inode(inode)->i_vino.snap;555}556 557/**558 * ceph_present_ino - format an inode number for presentation to userland559 * @sb: superblock where the inode lives560 * @ino: inode number to (possibly) convert561 *562 * If the user mounted with the ino32 option, then the 64-bit value needs563 * to be converted to something that can fit inside 32 bits. Note that564 * internal kernel code never uses this value, so this is entirely for565 * userland consumption.566 */567static inline u64 ceph_present_ino(struct super_block *sb, u64 ino)568{569 if (unlikely(ceph_test_mount_opt(ceph_sb_to_fs_client(sb), INO32)))570 return ceph_ino_to_ino32(ino);571 return ino;572}573 574static inline u64 ceph_present_inode(struct inode *inode)575{576 return ceph_present_ino(inode->i_sb, ceph_ino(inode));577}578 579static inline int ceph_ino_compare(struct inode *inode, void *data)580{581 struct ceph_vino *pvino = (struct ceph_vino *)data;582 struct ceph_inode_info *ci = ceph_inode(inode);583 return ci->i_vino.ino == pvino->ino &&584 ci->i_vino.snap == pvino->snap;585}586 587/*588 * The MDS reserves a set of inodes for its own usage. These should never589 * be accessible by clients, and so the MDS has no reason to ever hand these590 * out. The range is CEPH_MDS_INO_MDSDIR_OFFSET..CEPH_INO_SYSTEM_BASE.591 *592 * These come from src/mds/mdstypes.h in the ceph sources.593 */594#define CEPH_MAX_MDS 0x100595#define CEPH_NUM_STRAY 10596#define CEPH_MDS_INO_MDSDIR_OFFSET (1 * CEPH_MAX_MDS)597#define CEPH_MDS_INO_LOG_OFFSET (2 * CEPH_MAX_MDS)598#define CEPH_INO_SYSTEM_BASE ((6*CEPH_MAX_MDS) + (CEPH_MAX_MDS * CEPH_NUM_STRAY))599 600static inline bool ceph_vino_is_reserved(const struct ceph_vino vino)601{602 if (vino.ino >= CEPH_INO_SYSTEM_BASE ||603 vino.ino < CEPH_MDS_INO_MDSDIR_OFFSET)604 return false;605 606 /* Don't warn on mdsdirs */607 WARN_RATELIMIT(vino.ino >= CEPH_MDS_INO_LOG_OFFSET,608 "Attempt to access reserved inode number 0x%llx",609 vino.ino);610 return true;611}612 613static inline struct inode *ceph_find_inode(struct super_block *sb,614 struct ceph_vino vino)615{616 if (ceph_vino_is_reserved(vino))617 return NULL;618 619 /*620 * NB: The hashval will be run through the fs/inode.c hash function621 * anyway, so there is no need to squash the inode number down to622 * 32-bits first. Just use low-order bits on arches with 32-bit long.623 */624 return ilookup5(sb, (unsigned long)vino.ino, ceph_ino_compare, &vino);625}626 627 628/*629 * Ceph inode.630 */631#define CEPH_I_DIR_ORDERED (1 << 0) /* dentries in dir are ordered */632#define CEPH_I_FLUSH (1 << 2) /* do not delay flush of dirty metadata */633#define CEPH_I_POOL_PERM (1 << 3) /* pool rd/wr bits are valid */634#define CEPH_I_POOL_RD (1 << 4) /* can read from pool */635#define CEPH_I_POOL_WR (1 << 5) /* can write to pool */636#define CEPH_I_SEC_INITED (1 << 6) /* security initialized */637#define CEPH_I_KICK_FLUSH (1 << 7) /* kick flushing caps */638#define CEPH_I_FLUSH_SNAPS (1 << 8) /* need flush snapss */639#define CEPH_I_ERROR_WRITE (1 << 9) /* have seen write errors */640#define CEPH_I_ERROR_FILELOCK (1 << 10) /* have seen file lock errors */641#define CEPH_I_ODIRECT (1 << 11) /* inode in direct I/O mode */642#define CEPH_ASYNC_CREATE_BIT (12) /* async create in flight for this */643#define CEPH_I_ASYNC_CREATE (1 << CEPH_ASYNC_CREATE_BIT)644#define CEPH_I_SHUTDOWN (1 << 13) /* inode is no longer usable */645#define CEPH_I_ASYNC_CHECK_CAPS (1 << 14) /* check caps immediately after async646 creating finishes */647 648/*649 * Masks of ceph inode work.650 */651#define CEPH_I_WORK_WRITEBACK 0652#define CEPH_I_WORK_INVALIDATE_PAGES 1653#define CEPH_I_WORK_VMTRUNCATE 2654#define CEPH_I_WORK_CHECK_CAPS 3655#define CEPH_I_WORK_FLUSH_SNAPS 4656 657/*658 * We set the ERROR_WRITE bit when we start seeing write errors on an inode659 * and then clear it when they start succeeding. Note that we do a lockless660 * check first, and only take the lock if it looks like it needs to be changed.661 * The write submission code just takes this as a hint, so we're not too662 * worried if a few slip through in either direction.663 */664static inline void ceph_set_error_write(struct ceph_inode_info *ci)665{666 if (!(READ_ONCE(ci->i_ceph_flags) & CEPH_I_ERROR_WRITE)) {667 spin_lock(&ci->i_ceph_lock);668 ci->i_ceph_flags |= CEPH_I_ERROR_WRITE;669 spin_unlock(&ci->i_ceph_lock);670 }671}672 673static inline void ceph_clear_error_write(struct ceph_inode_info *ci)674{675 if (READ_ONCE(ci->i_ceph_flags) & CEPH_I_ERROR_WRITE) {676 spin_lock(&ci->i_ceph_lock);677 ci->i_ceph_flags &= ~CEPH_I_ERROR_WRITE;678 spin_unlock(&ci->i_ceph_lock);679 }680}681 682static inline void __ceph_dir_set_complete(struct ceph_inode_info *ci,683 long long release_count,684 long long ordered_count)685{686 /*687 * Makes sure operations that setup readdir cache (update page688 * cache and i_size) are strongly ordered w.r.t. the following689 * atomic64_set() operations.690 */691 smp_mb();692 atomic64_set(&ci->i_complete_seq[0], release_count);693 atomic64_set(&ci->i_complete_seq[1], ordered_count);694}695 696static inline void __ceph_dir_clear_complete(struct ceph_inode_info *ci)697{698 atomic64_inc(&ci->i_release_count);699}700 701static inline void __ceph_dir_clear_ordered(struct ceph_inode_info *ci)702{703 atomic64_inc(&ci->i_ordered_count);704}705 706static inline bool __ceph_dir_is_complete(struct ceph_inode_info *ci)707{708 return atomic64_read(&ci->i_complete_seq[0]) ==709 atomic64_read(&ci->i_release_count);710}711 712static inline bool __ceph_dir_is_complete_ordered(struct ceph_inode_info *ci)713{714 return atomic64_read(&ci->i_complete_seq[0]) ==715 atomic64_read(&ci->i_release_count) &&716 atomic64_read(&ci->i_complete_seq[1]) ==717 atomic64_read(&ci->i_ordered_count);718}719 720static inline void ceph_dir_clear_complete(struct inode *inode)721{722 __ceph_dir_clear_complete(ceph_inode(inode));723}724 725static inline void ceph_dir_clear_ordered(struct inode *inode)726{727 __ceph_dir_clear_ordered(ceph_inode(inode));728}729 730static inline bool ceph_dir_is_complete_ordered(struct inode *inode)731{732 bool ret = __ceph_dir_is_complete_ordered(ceph_inode(inode));733 smp_rmb();734 return ret;735}736 737/* find a specific frag @f */738extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci,739 u32 f);740 741/*742 * choose fragment for value @v. copy frag content to pfrag, if leaf743 * exists744 */745extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v,746 struct ceph_inode_frag *pfrag,747 int *found);748 749static inline struct ceph_dentry_info *ceph_dentry(const struct dentry *dentry)750{751 return (struct ceph_dentry_info *)dentry->d_fsdata;752}753 754/*755 * caps helpers756 */757static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci)758{759 return !RB_EMPTY_ROOT(&ci->i_caps);760}761 762extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented);763extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t);764extern int __ceph_caps_issued_mask_metric(struct ceph_inode_info *ci, int mask,765 int t);766extern int __ceph_caps_issued_other(struct ceph_inode_info *ci,767 struct ceph_cap *cap);768 769static inline int ceph_caps_issued(struct ceph_inode_info *ci)770{771 int issued;772 spin_lock(&ci->i_ceph_lock);773 issued = __ceph_caps_issued(ci, NULL);774 spin_unlock(&ci->i_ceph_lock);775 return issued;776}777 778static inline int ceph_caps_issued_mask_metric(struct ceph_inode_info *ci,779 int mask, int touch)780{781 int r;782 spin_lock(&ci->i_ceph_lock);783 r = __ceph_caps_issued_mask_metric(ci, mask, touch);784 spin_unlock(&ci->i_ceph_lock);785 return r;786}787 788static inline int __ceph_caps_dirty(struct ceph_inode_info *ci)789{790 return ci->i_dirty_caps | ci->i_flushing_caps;791}792extern struct ceph_cap_flush *ceph_alloc_cap_flush(void);793extern void ceph_free_cap_flush(struct ceph_cap_flush *cf);794extern int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask,795 struct ceph_cap_flush **pcf);796 797extern int __ceph_caps_revoking_other(struct ceph_inode_info *ci,798 struct ceph_cap *ocap, int mask);799extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask);800extern int __ceph_caps_used(struct ceph_inode_info *ci);801 802static inline bool __ceph_is_file_opened(struct ceph_inode_info *ci)803{804 return ci->i_nr_by_mode[0];805}806extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci);807extern int __ceph_caps_wanted(struct ceph_inode_info *ci);808 809/* what the mds thinks we want */810extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci, bool check);811 812extern void ceph_caps_init(struct ceph_mds_client *mdsc);813extern void ceph_caps_finalize(struct ceph_mds_client *mdsc);814extern void ceph_adjust_caps_max_min(struct ceph_mds_client *mdsc,815 struct ceph_mount_options *fsopt);816extern int ceph_reserve_caps(struct ceph_mds_client *mdsc,817 struct ceph_cap_reservation *ctx, int need);818extern void ceph_unreserve_caps(struct ceph_mds_client *mdsc,819 struct ceph_cap_reservation *ctx);820extern void ceph_reservation_status(struct ceph_fs_client *client,821 int *total, int *avail, int *used,822 int *reserved, int *min);823extern void change_auth_cap_ses(struct ceph_inode_info *ci,824 struct ceph_mds_session *session);825 826 827 828/*829 * we keep buffered readdir results attached to file->private_data830 */831#define CEPH_F_SYNC 1832#define CEPH_F_ATEND 2833 834struct ceph_file_info {835 short fmode; /* initialized on open */836 short flags; /* CEPH_F_* */837 838 spinlock_t rw_contexts_lock;839 struct list_head rw_contexts;840 841 u32 filp_gen;842};843 844struct ceph_dir_file_info {845 struct ceph_file_info file_info;846 847 /* readdir: position within the dir */848 u32 frag;849 struct ceph_mds_request *last_readdir;850 851 /* readdir: position within a frag */852 unsigned next_offset; /* offset of next chunk (last_name's + 1) */853 char *last_name; /* last entry in previous chunk */854 long long dir_release_count;855 long long dir_ordered_count;856 int readdir_cache_idx;857 858 /* used for -o dirstat read() on directory thing */859 char *dir_info;860 int dir_info_len;861};862 863struct ceph_rw_context {864 struct list_head list;865 struct task_struct *thread;866 int caps;867};868 869#define CEPH_DEFINE_RW_CONTEXT(_name, _caps) \870 struct ceph_rw_context _name = { \871 .thread = current, \872 .caps = _caps, \873 }874 875static inline void ceph_add_rw_context(struct ceph_file_info *cf,876 struct ceph_rw_context *ctx)877{878 spin_lock(&cf->rw_contexts_lock);879 list_add(&ctx->list, &cf->rw_contexts);880 spin_unlock(&cf->rw_contexts_lock);881}882 883static inline void ceph_del_rw_context(struct ceph_file_info *cf,884 struct ceph_rw_context *ctx)885{886 spin_lock(&cf->rw_contexts_lock);887 list_del(&ctx->list);888 spin_unlock(&cf->rw_contexts_lock);889}890 891static inline struct ceph_rw_context*892ceph_find_rw_context(struct ceph_file_info *cf)893{894 struct ceph_rw_context *ctx, *found = NULL;895 spin_lock(&cf->rw_contexts_lock);896 list_for_each_entry(ctx, &cf->rw_contexts, list) {897 if (ctx->thread == current) {898 found = ctx;899 break;900 }901 }902 spin_unlock(&cf->rw_contexts_lock);903 return found;904}905 906struct ceph_readdir_cache_control {907 struct page *page;908 struct dentry **dentries;909 int index;910};911 912/*913 * A "snap realm" describes a subset of the file hierarchy sharing914 * the same set of snapshots that apply to it. The realms themselves915 * are organized into a hierarchy, such that children inherit (some of)916 * the snapshots of their parents.917 *918 * All inodes within the realm that have capabilities are linked into a919 * per-realm list.920 */921struct ceph_snap_realm {922 u64 ino;923 struct inode *inode;924 atomic_t nref;925 struct rb_node node;926 927 u64 created, seq;928 u64 parent_ino;929 u64 parent_since; /* snapid when our current parent became so */930 931 u64 *prior_parent_snaps; /* snaps inherited from any parents we */932 u32 num_prior_parent_snaps; /* had prior to parent_since */933 u64 *snaps; /* snaps specific to this realm */934 u32 num_snaps;935 936 struct ceph_snap_realm *parent;937 struct list_head children; /* list of child realms */938 struct list_head child_item;939 940 struct list_head empty_item; /* if i have ref==0 */941 942 struct list_head dirty_item; /* if realm needs new context */943 944 struct list_head rebuild_item; /* rebuild snap realms _downward_ in hierarchy */945 946 /* the current set of snaps for this realm */947 struct ceph_snap_context *cached_context;948 949 struct list_head inodes_with_caps;950 spinlock_t inodes_with_caps_lock;951};952 953static inline int default_congestion_kb(void)954{955 int congestion_kb;956 957 /*958 * Copied from NFS959 *960 * congestion size, scale with available memory.961 *962 * 64MB: 8192k963 * 128MB: 11585k964 * 256MB: 16384k965 * 512MB: 23170k966 * 1GB: 32768k967 * 2GB: 46340k968 * 4GB: 65536k969 * 8GB: 92681k970 * 16GB: 131072k971 *972 * This allows larger machines to have larger/more transfers.973 * Limit the default to 256M974 */975 congestion_kb = (16*int_sqrt(totalram_pages())) << (PAGE_SHIFT-10);976 if (congestion_kb > 256*1024)977 congestion_kb = 256*1024;978 979 return congestion_kb;980}981 982 983/* super.c */984extern int ceph_force_reconnect(struct super_block *sb);985/* snap.c */986struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc,987 u64 ino);988extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc,989 struct ceph_snap_realm *realm);990extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc,991 struct ceph_snap_realm *realm);992extern int ceph_update_snap_trace(struct ceph_mds_client *m,993 void *p, void *e, bool deletion,994 struct ceph_snap_realm **realm_ret);995void ceph_change_snap_realm(struct inode *inode, struct ceph_snap_realm *realm);996extern void ceph_handle_snap(struct ceph_mds_client *mdsc,997 struct ceph_mds_session *session,998 struct ceph_msg *msg);999extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci,1000 struct ceph_cap_snap *capsnap);1001extern void ceph_cleanup_global_and_empty_realms(struct ceph_mds_client *mdsc);1002 1003extern struct ceph_snapid_map *ceph_get_snapid_map(struct ceph_mds_client *mdsc,1004 u64 snap);1005extern void ceph_put_snapid_map(struct ceph_mds_client* mdsc,1006 struct ceph_snapid_map *sm);1007extern void ceph_trim_snapid_map(struct ceph_mds_client *mdsc);1008extern void ceph_cleanup_snapid_map(struct ceph_mds_client *mdsc);1009void ceph_umount_begin(struct super_block *sb);1010 1011 1012/*1013 * a cap_snap is "pending" if it is still awaiting an in-progress1014 * sync write (that may/may not still update size, mtime, etc.).1015 */1016static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci)1017{1018 return !list_empty(&ci->i_cap_snaps) &&1019 list_last_entry(&ci->i_cap_snaps, struct ceph_cap_snap,1020 ci_item)->writing;1021}1022 1023/* inode.c */1024struct ceph_mds_reply_info_in;1025struct ceph_mds_reply_dirfrag;1026struct ceph_acl_sec_ctx;1027 1028extern const struct inode_operations ceph_file_iops;1029 1030extern struct inode *ceph_alloc_inode(struct super_block *sb);1031extern void ceph_evict_inode(struct inode *inode);1032extern void ceph_free_inode(struct inode *inode);1033 1034struct inode *ceph_new_inode(struct inode *dir, struct dentry *dentry,1035 umode_t *mode, struct ceph_acl_sec_ctx *as_ctx);1036void ceph_as_ctx_to_req(struct ceph_mds_request *req,1037 struct ceph_acl_sec_ctx *as_ctx);1038 1039extern struct inode *ceph_get_inode(struct super_block *sb,1040 struct ceph_vino vino,1041 struct inode *newino);1042extern struct inode *ceph_get_snapdir(struct inode *parent);1043extern int ceph_fill_file_size(struct inode *inode, int issued,1044 u32 truncate_seq, u64 truncate_size, u64 size);1045extern void ceph_fill_file_time(struct inode *inode, int issued,1046 u64 time_warp_seq, struct timespec64 *ctime,1047 struct timespec64 *mtime,1048 struct timespec64 *atime);1049extern int ceph_fill_inode(struct inode *inode, struct page *locked_page,1050 struct ceph_mds_reply_info_in *iinfo,1051 struct ceph_mds_reply_dirfrag *dirinfo,1052 struct ceph_mds_session *session, int cap_fmode,1053 struct ceph_cap_reservation *caps_reservation);1054extern int ceph_fill_trace(struct super_block *sb,1055 struct ceph_mds_request *req);1056extern int ceph_readdir_prepopulate(struct ceph_mds_request *req,1057 struct ceph_mds_session *session);1058 1059extern bool ceph_inode_set_size(struct inode *inode, loff_t size);1060extern void __ceph_do_pending_vmtruncate(struct inode *inode);1061 1062void ceph_queue_inode_work(struct inode *inode, int work_bit);1063 1064static inline void ceph_queue_vmtruncate(struct inode *inode)1065{1066 ceph_queue_inode_work(inode, CEPH_I_WORK_VMTRUNCATE);1067}1068 1069static inline void ceph_queue_invalidate(struct inode *inode)1070{1071 ceph_queue_inode_work(inode, CEPH_I_WORK_INVALIDATE_PAGES);1072}1073 1074static inline void ceph_queue_writeback(struct inode *inode)1075{1076 ceph_queue_inode_work(inode, CEPH_I_WORK_WRITEBACK);1077}1078 1079static inline void ceph_queue_check_caps(struct inode *inode)1080{1081 ceph_queue_inode_work(inode, CEPH_I_WORK_CHECK_CAPS);1082}1083 1084static inline void ceph_queue_flush_snaps(struct inode *inode)1085{1086 ceph_queue_inode_work(inode, CEPH_I_WORK_FLUSH_SNAPS);1087}1088 1089extern int ceph_try_to_choose_auth_mds(struct inode *inode, int mask);1090extern int __ceph_do_getattr(struct inode *inode, struct page *locked_page,1091 int mask, bool force);1092static inline int ceph_do_getattr(struct inode *inode, int mask, bool force)1093{1094 return __ceph_do_getattr(inode, NULL, mask, force);1095}1096extern int ceph_permission(struct mnt_idmap *idmap,1097 struct inode *inode, int mask);1098 1099struct ceph_iattr {1100 struct ceph_fscrypt_auth *fscrypt_auth;1101};1102 1103extern int __ceph_setattr(struct mnt_idmap *idmap, struct inode *inode,1104 struct iattr *attr, struct ceph_iattr *cia);1105extern int ceph_setattr(struct mnt_idmap *idmap,1106 struct dentry *dentry, struct iattr *attr);1107extern int ceph_getattr(struct mnt_idmap *idmap,1108 const struct path *path, struct kstat *stat,1109 u32 request_mask, unsigned int flags);1110void ceph_inode_shutdown(struct inode *inode);1111 1112static inline bool ceph_inode_is_shutdown(struct inode *inode)1113{1114 unsigned long flags = READ_ONCE(ceph_inode(inode)->i_ceph_flags);1115 struct ceph_fs_client *fsc = ceph_inode_to_fs_client(inode);1116 int state = READ_ONCE(fsc->mount_state);1117 1118 return (flags & CEPH_I_SHUTDOWN) || state >= CEPH_MOUNT_SHUTDOWN;1119}1120 1121/* xattr.c */1122int __ceph_setxattr(struct inode *, const char *, const void *, size_t, int);1123int ceph_do_getvxattr(struct inode *inode, const char *name, void *value, size_t size);1124ssize_t __ceph_getxattr(struct inode *, const char *, void *, size_t);1125extern ssize_t ceph_listxattr(struct dentry *, char *, size_t);1126extern struct ceph_buffer *__ceph_build_xattrs_blob(struct ceph_inode_info *ci);1127extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci);1128extern const struct xattr_handler * const ceph_xattr_handlers[];1129 1130struct ceph_acl_sec_ctx {1131#ifdef CONFIG_CEPH_FS_POSIX_ACL1132 void *default_acl;1133 void *acl;1134#endif1135#ifdef CONFIG_CEPH_FS_SECURITY_LABEL1136 void *sec_ctx;1137 u32 sec_ctxlen;1138#endif1139#ifdef CONFIG_FS_ENCRYPTION1140 struct ceph_fscrypt_auth *fscrypt_auth;1141#endif1142 struct ceph_pagelist *pagelist;1143};1144 1145#ifdef CONFIG_SECURITY1146extern bool ceph_security_xattr_deadlock(struct inode *in);1147extern bool ceph_security_xattr_wanted(struct inode *in);1148#else1149static inline bool ceph_security_xattr_deadlock(struct inode *in)1150{1151 return false;1152}1153static inline bool ceph_security_xattr_wanted(struct inode *in)1154{1155 return false;1156}1157#endif1158 1159#ifdef CONFIG_CEPH_FS_SECURITY_LABEL1160extern int ceph_security_init_secctx(struct dentry *dentry, umode_t mode,1161 struct ceph_acl_sec_ctx *ctx);1162static inline void ceph_security_invalidate_secctx(struct inode *inode)1163{1164 security_inode_invalidate_secctx(inode);1165}1166#else1167static inline int ceph_security_init_secctx(struct dentry *dentry, umode_t mode,1168 struct ceph_acl_sec_ctx *ctx)1169{1170 return 0;1171}1172static inline void ceph_security_invalidate_secctx(struct inode *inode)1173{1174}1175#endif1176 1177void ceph_release_acl_sec_ctx(struct ceph_acl_sec_ctx *as_ctx);1178 1179/* acl.c */1180#ifdef CONFIG_CEPH_FS_POSIX_ACL1181 1182struct posix_acl *ceph_get_acl(struct inode *, int, bool);1183int ceph_set_acl(struct mnt_idmap *idmap,1184 struct dentry *dentry, struct posix_acl *acl, int type);1185int ceph_pre_init_acls(struct inode *dir, umode_t *mode,1186 struct ceph_acl_sec_ctx *as_ctx);1187void ceph_init_inode_acls(struct inode *inode,1188 struct ceph_acl_sec_ctx *as_ctx);1189 1190static inline void ceph_forget_all_cached_acls(struct inode *inode)1191{1192 forget_all_cached_acls(inode);1193}1194 1195#else1196 1197#define ceph_get_acl NULL1198#define ceph_set_acl NULL1199 1200static inline int ceph_pre_init_acls(struct inode *dir, umode_t *mode,1201 struct ceph_acl_sec_ctx *as_ctx)1202{1203 return 0;1204}1205static inline void ceph_init_inode_acls(struct inode *inode,1206 struct ceph_acl_sec_ctx *as_ctx)1207{1208}1209 1210static inline void ceph_forget_all_cached_acls(struct inode *inode)1211{1212}1213 1214#endif1215 1216/* caps.c */1217extern const char *ceph_cap_string(int c);1218extern void ceph_handle_caps(struct ceph_mds_session *session,1219 struct ceph_msg *msg);1220extern struct ceph_cap *ceph_get_cap(struct ceph_mds_client *mdsc,1221 struct ceph_cap_reservation *ctx);1222extern void ceph_add_cap(struct inode *inode,1223 struct ceph_mds_session *session, u64 cap_id,1224 unsigned issued, unsigned wanted,1225 unsigned cap, unsigned seq, u64 realmino, int flags,1226 struct ceph_cap **new_cap);1227extern void __ceph_remove_cap(struct ceph_cap *cap, bool queue_release);1228extern void ceph_remove_cap(struct ceph_mds_client *mdsc, struct ceph_cap *cap,1229 bool queue_release);1230extern void __ceph_remove_caps(struct ceph_inode_info *ci);1231extern void ceph_put_cap(struct ceph_mds_client *mdsc,1232 struct ceph_cap *cap);1233extern int ceph_is_any_caps(struct inode *inode);1234 1235extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc);1236extern int ceph_fsync(struct file *file, loff_t start, loff_t end,1237 int datasync);1238extern void ceph_early_kick_flushing_caps(struct ceph_mds_client *mdsc,1239 struct ceph_mds_session *session);1240extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc,1241 struct ceph_mds_session *session);1242void ceph_kick_flushing_inode_caps(struct ceph_mds_session *session,1243 struct ceph_inode_info *ci);1244extern struct ceph_cap *__get_cap_for_mds(struct ceph_inode_info *ci,1245 int mds);1246extern struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci,1247 int mds);1248extern void ceph_take_cap_refs(struct ceph_inode_info *ci, int caps,1249 bool snap_rwsem_locked);1250extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps);1251extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had);1252extern void ceph_put_cap_refs_async(struct ceph_inode_info *ci, int had);1253extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,1254 struct ceph_snap_context *snapc);1255extern void __ceph_remove_capsnap(struct inode *inode,1256 struct ceph_cap_snap *capsnap,1257 bool *wake_ci, bool *wake_mdsc);1258extern void ceph_remove_capsnap(struct inode *inode,1259 struct ceph_cap_snap *capsnap,1260 bool *wake_ci, bool *wake_mdsc);1261extern void ceph_flush_snaps(struct ceph_inode_info *ci,1262 struct ceph_mds_session **psession);1263extern bool __ceph_should_report_size(struct ceph_inode_info *ci);1264extern void ceph_check_caps(struct ceph_inode_info *ci, int flags);1265extern unsigned long ceph_check_delayed_caps(struct ceph_mds_client *mdsc);1266extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc);1267extern void ceph_flush_cap_releases(struct ceph_mds_client *mdsc);1268extern int ceph_drop_caps_for_unlink(struct inode *inode);1269extern int ceph_encode_inode_release(void **p, struct inode *inode,1270 int mds, int drop, int unless, int force);1271extern int ceph_encode_dentry_release(void **p, struct dentry *dn,1272 struct inode *dir,1273 int mds, int drop, int unless);1274 1275extern int __ceph_get_caps(struct inode *inode, struct ceph_file_info *fi,1276 int need, int want, loff_t endoff, int *got);1277extern int ceph_get_caps(struct file *filp, int need, int want,1278 loff_t endoff, int *got);1279extern int ceph_try_get_caps(struct inode *inode,1280 int need, int want, bool nonblock, int *got);1281 1282/* for counting open files by mode */1283extern void ceph_get_fmode(struct ceph_inode_info *ci, int mode, int count);1284extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode, int count);1285extern void __ceph_touch_fmode(struct ceph_inode_info *ci,1286 struct ceph_mds_client *mdsc, int fmode);1287 1288/* addr.c */1289extern const struct address_space_operations ceph_aops;1290extern const struct netfs_request_ops ceph_netfs_ops;1291extern int ceph_mmap(struct file *file, struct vm_area_struct *vma);1292extern int ceph_uninline_data(struct file *file);1293extern int ceph_pool_perm_check(struct inode *inode, int need);1294extern void ceph_pool_perm_destroy(struct ceph_mds_client* mdsc);1295int ceph_purge_inode_cap(struct inode *inode, struct ceph_cap *cap, bool *invalidate);1296 1297static inline bool ceph_has_inline_data(struct ceph_inode_info *ci)1298{1299 if (ci->i_inline_version == CEPH_INLINE_NONE ||1300 ci->i_inline_version == 1) /* initial version, no data */1301 return false;1302 return true;1303}1304 1305/* file.c */1306extern const struct file_operations ceph_file_fops;1307 1308extern int ceph_renew_caps(struct inode *inode, int fmode);1309extern int ceph_open(struct inode *inode, struct file *file);1310extern int ceph_atomic_open(struct inode *dir, struct dentry *dentry,1311 struct file *file, unsigned flags, umode_t mode);1312extern ssize_t __ceph_sync_read(struct inode *inode, loff_t *ki_pos,1313 struct iov_iter *to, int *retry_op,1314 u64 *last_objver);1315extern int ceph_release(struct inode *inode, struct file *filp);1316extern void ceph_fill_inline_data(struct inode *inode, struct page *locked_page,1317 char *data, size_t len);1318 1319/* dir.c */1320extern const struct file_operations ceph_dir_fops;1321extern const struct file_operations ceph_snapdir_fops;1322extern const struct inode_operations ceph_dir_iops;1323extern const struct inode_operations ceph_snapdir_iops;1324extern const struct dentry_operations ceph_dentry_ops;1325 1326extern loff_t ceph_make_fpos(unsigned high, unsigned off, bool hash_order);1327extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry);1328extern struct dentry *ceph_handle_snapdir(struct ceph_mds_request *req,1329 struct dentry *dentry);1330extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req,1331 struct dentry *dentry, int err);1332 1333extern void __ceph_dentry_lease_touch(struct ceph_dentry_info *di);1334extern void __ceph_dentry_dir_lease_touch(struct ceph_dentry_info *di);1335extern void ceph_invalidate_dentry_lease(struct dentry *dentry);1336extern int ceph_trim_dentries(struct ceph_mds_client *mdsc);1337extern unsigned ceph_dentry_hash(struct inode *dir, struct dentry *dn);1338extern void ceph_readdir_cache_release(struct ceph_readdir_cache_control *ctl);1339 1340/* ioctl.c */1341extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg);1342 1343/* export.c */1344extern const struct export_operations ceph_export_ops;1345struct inode *ceph_lookup_inode(struct super_block *sb, u64 ino);1346 1347/* locks.c */1348extern __init void ceph_flock_init(void);1349extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl);1350extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl);1351extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num);1352extern int ceph_encode_locks_to_buffer(struct inode *inode,1353 struct ceph_filelock *flocks,1354 int num_fcntl_locks,1355 int num_flock_locks);1356extern int ceph_locks_to_pagelist(struct ceph_filelock *flocks,1357 struct ceph_pagelist *pagelist,1358 int num_fcntl_locks, int num_flock_locks);1359 1360/* debugfs.c */1361extern void ceph_fs_debugfs_init(struct ceph_fs_client *client);1362extern void ceph_fs_debugfs_cleanup(struct ceph_fs_client *client);1363 1364/* quota.c */1365 1366enum quota_get_realm {1367 QUOTA_GET_MAX_FILES,1368 QUOTA_GET_MAX_BYTES,1369 QUOTA_GET_ANY1370};1371 1372static inline bool __ceph_has_quota(struct ceph_inode_info *ci,1373 enum quota_get_realm which)1374{1375 bool has_quota = false;1376 1377 switch (which) {1378 case QUOTA_GET_MAX_BYTES:1379 has_quota = !!ci->i_max_bytes;1380 break;1381 case QUOTA_GET_MAX_FILES:1382 has_quota = !!ci->i_max_files;1383 break;1384 default:1385 has_quota = !!(ci->i_max_files || ci->i_max_bytes);1386 }1387 return has_quota;1388}1389 1390extern void ceph_adjust_quota_realms_count(struct inode *inode, bool inc);1391 1392static inline void __ceph_update_quota(struct ceph_inode_info *ci,1393 u64 max_bytes, u64 max_files)1394{1395 bool had_quota, has_quota;1396 had_quota = __ceph_has_quota(ci, QUOTA_GET_ANY);1397 ci->i_max_bytes = max_bytes;1398 ci->i_max_files = max_files;1399 has_quota = __ceph_has_quota(ci, QUOTA_GET_ANY);1400 1401 if (had_quota != has_quota)1402 ceph_adjust_quota_realms_count(&ci->netfs.inode, has_quota);1403}1404 1405static inline int __ceph_sparse_read_ext_count(struct inode *inode, u64 len)1406{1407 int cnt = 0;1408 1409 if (IS_ENCRYPTED(inode)) {1410 cnt = len >> CEPH_FSCRYPT_BLOCK_SHIFT;1411 if (cnt > CEPH_SPARSE_EXT_ARRAY_INITIAL)1412 cnt = 0;1413 }1414 1415 return cnt;1416}1417 1418extern void ceph_handle_quota(struct ceph_mds_client *mdsc,1419 struct ceph_mds_session *session,1420 struct ceph_msg *msg);1421extern bool ceph_quota_is_max_files_exceeded(struct inode *inode);1422extern bool ceph_quota_is_same_realm(struct inode *old, struct inode *new);1423extern bool ceph_quota_is_max_bytes_exceeded(struct inode *inode,1424 loff_t newlen);1425extern bool ceph_quota_is_max_bytes_approaching(struct inode *inode,1426 loff_t newlen);1427extern bool ceph_quota_update_statfs(struct ceph_fs_client *fsc,1428 struct kstatfs *buf);1429extern void ceph_cleanup_quotarealms_inodes(struct ceph_mds_client *mdsc);1430 1431bool ceph_inc_mds_stopping_blocker(struct ceph_mds_client *mdsc,1432 struct ceph_mds_session *session);1433void ceph_dec_mds_stopping_blocker(struct ceph_mds_client *mdsc);1434bool ceph_inc_osd_stopping_blocker(struct ceph_mds_client *mdsc);1435void ceph_dec_osd_stopping_blocker(struct ceph_mds_client *mdsc);1436#endif /* _FS_CEPH_SUPER_H */1437