1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef _FS_CEPH_SUPER_H
3 #define _FS_CEPH_SUPER_H
5 #include <linux/ceph/ceph_debug.h>
7 #include <asm/unaligned.h>
8 #include <linux/backing-dev.h>
9 #include <linux/completion.h>
10 #include <linux/exportfs.h>
12 #include <linux/mempool.h>
13 #include <linux/pagemap.h>
14 #include <linux/wait.h>
15 #include <linux/writeback.h>
16 #include <linux/slab.h>
17 #include <linux/posix_acl.h>
18 #include <linux/refcount.h>
19 #include <linux/security.h>
21 #include <linux/ceph/libceph.h>
23 #ifdef CONFIG_CEPH_FSCACHE
24 #include <linux/fscache.h>
27 /* f_type in struct statfs */
28 #define CEPH_SUPER_MAGIC 0x00c36400
30 /* large granularity for statfs utilization stats to facilitate
31 * large volume sizes on 32-bit machines. */
32 #define CEPH_BLOCK_SHIFT 22 /* 4 MB */
33 #define CEPH_BLOCK (1 << CEPH_BLOCK_SHIFT)
35 #define CEPH_MOUNT_OPT_CLEANRECOVER (1<<1) /* auto reonnect (clean mode) after blocklisted */
36 #define CEPH_MOUNT_OPT_DIRSTAT (1<<4) /* `cat dirname` for stats */
37 #define CEPH_MOUNT_OPT_RBYTES (1<<5) /* dir st_bytes = rbytes */
38 #define CEPH_MOUNT_OPT_NOASYNCREADDIR (1<<7) /* no dcache readdir */
39 #define CEPH_MOUNT_OPT_INO32 (1<<8) /* 32 bit inos */
40 #define CEPH_MOUNT_OPT_DCACHE (1<<9) /* use dcache for readdir etc */
41 #define CEPH_MOUNT_OPT_FSCACHE (1<<10) /* use fscache */
42 #define CEPH_MOUNT_OPT_NOPOOLPERM (1<<11) /* no pool permission check */
43 #define CEPH_MOUNT_OPT_MOUNTWAIT (1<<12) /* mount waits if no mds is up */
44 #define CEPH_MOUNT_OPT_NOQUOTADF (1<<13) /* no root dir quota in statfs */
45 #define CEPH_MOUNT_OPT_NOCOPYFROM (1<<14) /* don't use RADOS 'copy-from' op */
46 #define CEPH_MOUNT_OPT_ASYNC_DIROPS (1<<15) /* allow async directory ops */
48 #define CEPH_MOUNT_OPT_DEFAULT \
49 (CEPH_MOUNT_OPT_DCACHE | \
50 CEPH_MOUNT_OPT_NOCOPYFROM)
52 #define ceph_set_mount_opt(fsc, opt) \
53 (fsc)->mount_options->flags |= CEPH_MOUNT_OPT_##opt
54 #define ceph_clear_mount_opt(fsc, opt) \
55 (fsc)->mount_options->flags &= ~CEPH_MOUNT_OPT_##opt
56 #define ceph_test_mount_opt(fsc, opt) \
57 (!!((fsc)->mount_options->flags & CEPH_MOUNT_OPT_##opt))
59 /* max size of osd read request, limited by libceph */
60 #define CEPH_MAX_READ_SIZE CEPH_MSG_MAX_DATA_LEN
61 /* osd has a configurable limitaion of max write size.
62 * CEPH_MSG_MAX_DATA_LEN should be small enough. */
63 #define CEPH_MAX_WRITE_SIZE CEPH_MSG_MAX_DATA_LEN
64 #define CEPH_RASIZE_DEFAULT (8192*1024) /* max readahead */
65 #define CEPH_MAX_READDIR_DEFAULT 1024
66 #define CEPH_MAX_READDIR_BYTES_DEFAULT (512*1024)
67 #define CEPH_SNAPDIRNAME_DEFAULT ".snap"
70 * Delay telling the MDS we no longer want caps, in case we reopen
71 * the file. Delay a minimum amount of time, even if we send a cap
72 * message for some other reason. Otherwise, take the oppotunity to
73 * update the mds to avoid sending another message later.
75 #define CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT 5 /* cap release delay */
76 #define CEPH_CAPS_WANTED_DELAY_MAX_DEFAULT 60 /* cap release delay */
78 struct ceph_mount_options {
81 unsigned int wsize; /* max write size */
82 unsigned int rsize; /* max read size */
83 unsigned int rasize; /* max readahead */
84 unsigned int congestion_kb; /* max writeback in flight */
85 unsigned int caps_wanted_delay_min, caps_wanted_delay_max;
87 unsigned int max_readdir; /* max readdir result (entries) */
88 unsigned int max_readdir_bytes; /* max readdir result (bytes) */
91 * everything above this point can be memcmp'd; everything below
92 * is handled in compare_mount_options()
95 char *snapdir_name; /* default ".snap" */
96 char *mds_namespace; /* default NULL */
97 char *server_path; /* default NULL (means "/") */
98 char *fscache_uniq; /* default NULL */
101 struct ceph_fs_client {
102 struct super_block *sb;
104 struct list_head metric_wakeup;
106 struct ceph_mount_options *mount_options;
107 struct ceph_client *client;
109 unsigned long mount_state;
111 unsigned long last_auto_reconnect;
114 bool have_copy_from2;
117 loff_t max_file_size;
119 struct ceph_mds_client *mdsc;
121 atomic_long_t writeback_count;
123 struct workqueue_struct *inode_wq;
124 struct workqueue_struct *cap_wq;
126 #ifdef CONFIG_DEBUG_FS
127 struct dentry *debugfs_dentry_lru, *debugfs_caps;
128 struct dentry *debugfs_congestion_kb;
129 struct dentry *debugfs_bdi;
130 struct dentry *debugfs_mdsc, *debugfs_mdsmap;
131 struct dentry *debugfs_metric;
132 struct dentry *debugfs_mds_sessions;
135 #ifdef CONFIG_CEPH_FSCACHE
136 struct fscache_cookie *fscache;
142 * File i/o capability. This tracks shared state with the metadata
143 * server that allows us to cache or writeback attributes or to read
144 * and write data. For any given inode, we should have one or more
145 * capabilities, one issued by each metadata server, and our
146 * cumulative access is the OR of all issued capabilities.
148 * Each cap is referenced by the inode's i_caps rbtree and by per-mds
149 * session capability lists.
152 struct ceph_inode_info *ci;
153 struct rb_node ci_node; /* per-ci cap tree */
154 struct ceph_mds_session *session;
155 struct list_head session_caps; /* per-session caplist */
156 u64 cap_id; /* unique cap id (mds provided) */
160 int issued; /* latest, from the mds */
161 int implemented; /* implemented superset of
162 issued (for revocation) */
165 /* caps to release */
171 u32 seq, issue_seq, mseq;
172 u32 cap_gen; /* active/stale cycle */
173 unsigned long last_used;
174 struct list_head caps_item;
177 #define CHECK_CAPS_AUTHONLY 1 /* only check auth cap */
178 #define CHECK_CAPS_FLUSH 2 /* flush any dirty caps */
179 #define CHECK_CAPS_NOINVAL 4 /* don't invalidate pagecache */
181 struct ceph_cap_flush {
183 int caps; /* 0 means capsnap */
184 bool wake; /* wake up flush waiters when finish ? */
185 struct list_head g_list; // global
186 struct list_head i_list; // per inode
190 * Snapped cap state that is pending flush to mds. When a snapshot occurs,
191 * we first complete any in-process sync writes and writeback any dirty
192 * data before flushing the snapped state (tracked here) back to the MDS.
194 struct ceph_cap_snap {
196 struct list_head ci_item;
198 struct ceph_cap_flush cap_flush;
202 struct ceph_snap_context *context;
208 struct ceph_buffer *xattr_blob;
213 struct timespec64 mtime, atime, ctime, btime;
217 int writing; /* a sync write is still in progress */
218 int dirty_pages; /* dirty pages awaiting writeback */
223 static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap)
225 if (refcount_dec_and_test(&capsnap->nref)) {
226 if (capsnap->xattr_blob)
227 ceph_buffer_put(capsnap->xattr_blob);
233 * The frag tree describes how a directory is fragmented, potentially across
234 * multiple metadata servers. It is also used to indicate points where
235 * metadata authority is delegated, and whether/where metadata is replicated.
237 * A _leaf_ frag will be present in the i_fragtree IFF there is
238 * delegation info. That is, if mds >= 0 || ndist > 0.
240 #define CEPH_MAX_DIRFRAG_REP 4
242 struct ceph_inode_frag {
247 int split_by; /* i.e. 2^(split_by) children */
249 /* delegation and replication info */
250 int mds; /* -1 if same authority as parent */
251 int ndist; /* >0 if replicated */
252 int dist[CEPH_MAX_DIRFRAG_REP];
256 * We cache inode xattrs as an encoded blob until they are first used,
257 * at which point we parse them into an rbtree.
259 struct ceph_inode_xattr {
268 int should_free_name;
275 struct ceph_dentry_info {
276 struct dentry *dentry;
277 struct ceph_mds_session *lease_session;
278 struct list_head lease_list;
280 int lease_shared_gen;
283 unsigned long lease_renew_after, lease_renew_from;
288 #define CEPH_DENTRY_REFERENCED 1
289 #define CEPH_DENTRY_LEASE_LIST 2
290 #define CEPH_DENTRY_SHRINK_LIST 4
291 #define CEPH_DENTRY_PRIMARY_LINK 8
293 struct ceph_inode_xattrs_info {
295 * (still encoded) xattr blob. we avoid the overhead of parsing
296 * this until someone actually calls getxattr, etc.
298 * blob->vec.iov_len == 4 implies there are no xattrs; blob ==
299 * NULL means we don't know.
301 struct ceph_buffer *blob, *prealloc_blob;
303 struct rb_root index;
308 u64 version, index_version;
314 struct ceph_inode_info {
315 struct ceph_vino i_vino; /* ceph ino + snap */
317 spinlock_t i_ceph_lock;
320 u64 i_inline_version;
323 unsigned long i_ceph_flags;
324 atomic64_t i_release_count;
325 atomic64_t i_ordered_count;
326 atomic64_t i_complete_seq[2];
328 struct ceph_dir_layout i_dir_layout;
329 struct ceph_file_layout i_layout;
330 struct ceph_file_layout i_cached_layout; // for async creates
334 struct timespec64 i_rctime;
335 u64 i_rbytes, i_rfiles, i_rsubdirs;
336 u64 i_files, i_subdirs;
339 u64 i_max_bytes, i_max_files;
343 struct rb_root i_fragtree;
344 int i_fragtree_nsplits;
345 struct mutex i_fragtree_mutex;
347 struct ceph_inode_xattrs_info i_xattrs;
349 /* capabilities. protected _both_ by i_ceph_lock and cap->session's
351 struct rb_root i_caps; /* cap list */
352 struct ceph_cap *i_auth_cap; /* authoritative cap, if any */
353 unsigned i_dirty_caps, i_flushing_caps; /* mask of dirtied fields */
356 * Link to the auth cap's session's s_cap_dirty list. s_cap_dirty
357 * is protected by the mdsc->cap_dirty_lock, but each individual item
358 * is also protected by the inode's i_ceph_lock. Walking s_cap_dirty
359 * requires the mdsc->cap_dirty_lock. List presence for an item can
360 * be tested under the i_ceph_lock. Changing anything requires both.
362 struct list_head i_dirty_item;
365 * Link to session's s_cap_flushing list. Protected in a similar
366 * fashion to i_dirty_item, but also by the s_mutex for changes. The
367 * s_cap_flushing list can be walked while holding either the s_mutex
368 * or msdc->cap_dirty_lock. List presence can also be checked while
369 * holding the i_ceph_lock for this inode.
371 struct list_head i_flushing_item;
373 /* we need to track cap writeback on a per-cap-bit basis, to allow
374 * overlapping, pipelined cap flushes to the mds. we can probably
375 * reduce the tid to 8 bits if we're concerned about inode size. */
376 struct ceph_cap_flush *i_prealloc_cap_flush;
377 struct list_head i_cap_flush_list;
378 wait_queue_head_t i_cap_wq; /* threads waiting on a capability */
379 unsigned long i_hold_caps_max; /* jiffies */
380 struct list_head i_cap_delay_list; /* for delayed cap release to mds */
381 struct ceph_cap_reservation i_cap_migration_resv;
382 struct list_head i_cap_snaps; /* snapped state pending flush to mds */
383 struct ceph_snap_context *i_head_snapc; /* set if wr_buffer_head > 0 or
384 dirty|flushing caps */
385 unsigned i_snap_caps; /* cap bits for snapped files */
387 unsigned long i_last_rd;
388 unsigned long i_last_wr;
389 int i_nr_by_mode[CEPH_FILE_MODE_BITS]; /* open file counts */
391 struct mutex i_truncate_mutex;
392 u32 i_truncate_seq; /* last truncate to smaller size */
393 u64 i_truncate_size; /* and the size we last truncated down to */
394 int i_truncate_pending; /* still need to call vmtruncate */
396 u64 i_max_size; /* max file size authorized by mds */
397 u64 i_reported_size; /* (max_)size reported to or requested of mds */
398 u64 i_wanted_max_size; /* offset we'd like to write too */
399 u64 i_requested_max_size; /* max_size we've requested */
401 /* held references to caps */
403 int i_rd_ref, i_rdcache_ref, i_wr_ref, i_wb_ref, i_fx_ref;
404 int i_wrbuffer_ref, i_wrbuffer_ref_head;
405 atomic_t i_filelock_ref;
406 atomic_t i_shared_gen; /* increment each time we get FILE_SHARED */
407 u32 i_rdcache_gen; /* incremented each time we get FILE_CACHE. */
408 u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */
410 struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */
411 struct list_head i_unsafe_iops; /* uncommitted mds inode ops */
412 spinlock_t i_unsafe_lock;
415 struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */
416 struct ceph_snapid_map *i_snapid_map; /* snapid -> dev_t */
418 int i_snap_realm_counter; /* snap realm (if caps) */
419 struct list_head i_snap_realm_item;
420 struct list_head i_snap_flush_item;
421 struct timespec64 i_btime;
422 struct timespec64 i_snap_btime;
424 struct work_struct i_work;
425 unsigned long i_work_mask;
427 #ifdef CONFIG_CEPH_FSCACHE
428 struct fscache_cookie *fscache;
433 struct inode vfs_inode; /* at end */
436 static inline struct ceph_inode_info *
437 ceph_inode(const struct inode *inode)
439 return container_of(inode, struct ceph_inode_info, vfs_inode);
442 static inline struct ceph_fs_client *
443 ceph_inode_to_client(const struct inode *inode)
445 return (struct ceph_fs_client *)inode->i_sb->s_fs_info;
448 static inline struct ceph_fs_client *
449 ceph_sb_to_client(const struct super_block *sb)
451 return (struct ceph_fs_client *)sb->s_fs_info;
454 static inline struct ceph_mds_client *
455 ceph_sb_to_mdsc(const struct super_block *sb)
457 return (struct ceph_mds_client *)ceph_sb_to_client(sb)->mdsc;
460 static inline struct ceph_vino
461 ceph_vino(const struct inode *inode)
463 return ceph_inode(inode)->i_vino;
466 static inline u32 ceph_ino_to_ino32(u64 vino)
468 u32 ino = vino & 0xffffffff;
476 * Inode numbers in cephfs are 64 bits, but inode->i_ino is 32-bits on
477 * some arches. We generally do not use this value inside the ceph driver, but
478 * we do want to set it to something, so that generic vfs code has an
479 * appropriate value for tracepoints and the like.
481 static inline ino_t ceph_vino_to_ino_t(struct ceph_vino vino)
483 if (sizeof(ino_t) == sizeof(u32))
484 return ceph_ino_to_ino32(vino.ino);
485 return (ino_t)vino.ino;
488 /* for printf-style formatting */
489 #define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap
491 static inline u64 ceph_ino(struct inode *inode)
493 return ceph_inode(inode)->i_vino.ino;
496 static inline u64 ceph_snap(struct inode *inode)
498 return ceph_inode(inode)->i_vino.snap;
502 * ceph_present_ino - format an inode number for presentation to userland
503 * @sb: superblock where the inode lives
504 * @ino: inode number to (possibly) convert
506 * If the user mounted with the ino32 option, then the 64-bit value needs
507 * to be converted to something that can fit inside 32 bits. Note that
508 * internal kernel code never uses this value, so this is entirely for
509 * userland consumption.
511 static inline u64 ceph_present_ino(struct super_block *sb, u64 ino)
513 if (unlikely(ceph_test_mount_opt(ceph_sb_to_client(sb), INO32)))
514 return ceph_ino_to_ino32(ino);
518 static inline u64 ceph_present_inode(struct inode *inode)
520 return ceph_present_ino(inode->i_sb, ceph_ino(inode));
523 static inline int ceph_ino_compare(struct inode *inode, void *data)
525 struct ceph_vino *pvino = (struct ceph_vino *)data;
526 struct ceph_inode_info *ci = ceph_inode(inode);
527 return ci->i_vino.ino == pvino->ino &&
528 ci->i_vino.snap == pvino->snap;
532 static inline struct inode *ceph_find_inode(struct super_block *sb,
533 struct ceph_vino vino)
536 * NB: The hashval will be run through the fs/inode.c hash function
537 * anyway, so there is no need to squash the inode number down to
538 * 32-bits first. Just use low-order bits on arches with 32-bit long.
540 return ilookup5(sb, (unsigned long)vino.ino, ceph_ino_compare, &vino);
547 #define CEPH_I_DIR_ORDERED (1 << 0) /* dentries in dir are ordered */
548 #define CEPH_I_FLUSH (1 << 2) /* do not delay flush of dirty metadata */
549 #define CEPH_I_POOL_PERM (1 << 3) /* pool rd/wr bits are valid */
550 #define CEPH_I_POOL_RD (1 << 4) /* can read from pool */
551 #define CEPH_I_POOL_WR (1 << 5) /* can write to pool */
552 #define CEPH_I_SEC_INITED (1 << 6) /* security initialized */
553 #define CEPH_I_KICK_FLUSH (1 << 7) /* kick flushing caps */
554 #define CEPH_I_FLUSH_SNAPS (1 << 8) /* need flush snapss */
555 #define CEPH_I_ERROR_WRITE (1 << 9) /* have seen write errors */
556 #define CEPH_I_ERROR_FILELOCK (1 << 10) /* have seen file lock errors */
557 #define CEPH_I_ODIRECT (1 << 11) /* inode in direct I/O mode */
558 #define CEPH_ASYNC_CREATE_BIT (12) /* async create in flight for this */
559 #define CEPH_I_ASYNC_CREATE (1 << CEPH_ASYNC_CREATE_BIT)
562 * Masks of ceph inode work.
564 #define CEPH_I_WORK_WRITEBACK 0 /* writeback */
565 #define CEPH_I_WORK_INVALIDATE_PAGES 1 /* invalidate pages */
566 #define CEPH_I_WORK_VMTRUNCATE 2 /* vmtruncate */
569 * We set the ERROR_WRITE bit when we start seeing write errors on an inode
570 * and then clear it when they start succeeding. Note that we do a lockless
571 * check first, and only take the lock if it looks like it needs to be changed.
572 * The write submission code just takes this as a hint, so we're not too
573 * worried if a few slip through in either direction.
575 static inline void ceph_set_error_write(struct ceph_inode_info *ci)
577 if (!(READ_ONCE(ci->i_ceph_flags) & CEPH_I_ERROR_WRITE)) {
578 spin_lock(&ci->i_ceph_lock);
579 ci->i_ceph_flags |= CEPH_I_ERROR_WRITE;
580 spin_unlock(&ci->i_ceph_lock);
584 static inline void ceph_clear_error_write(struct ceph_inode_info *ci)
586 if (READ_ONCE(ci->i_ceph_flags) & CEPH_I_ERROR_WRITE) {
587 spin_lock(&ci->i_ceph_lock);
588 ci->i_ceph_flags &= ~CEPH_I_ERROR_WRITE;
589 spin_unlock(&ci->i_ceph_lock);
593 static inline void __ceph_dir_set_complete(struct ceph_inode_info *ci,
594 long long release_count,
595 long long ordered_count)
598 * Makes sure operations that setup readdir cache (update page
599 * cache and i_size) are strongly ordered w.r.t. the following
600 * atomic64_set() operations.
603 atomic64_set(&ci->i_complete_seq[0], release_count);
604 atomic64_set(&ci->i_complete_seq[1], ordered_count);
607 static inline void __ceph_dir_clear_complete(struct ceph_inode_info *ci)
609 atomic64_inc(&ci->i_release_count);
612 static inline void __ceph_dir_clear_ordered(struct ceph_inode_info *ci)
614 atomic64_inc(&ci->i_ordered_count);
617 static inline bool __ceph_dir_is_complete(struct ceph_inode_info *ci)
619 return atomic64_read(&ci->i_complete_seq[0]) ==
620 atomic64_read(&ci->i_release_count);
623 static inline bool __ceph_dir_is_complete_ordered(struct ceph_inode_info *ci)
625 return atomic64_read(&ci->i_complete_seq[0]) ==
626 atomic64_read(&ci->i_release_count) &&
627 atomic64_read(&ci->i_complete_seq[1]) ==
628 atomic64_read(&ci->i_ordered_count);
631 static inline void ceph_dir_clear_complete(struct inode *inode)
633 __ceph_dir_clear_complete(ceph_inode(inode));
636 static inline void ceph_dir_clear_ordered(struct inode *inode)
638 __ceph_dir_clear_ordered(ceph_inode(inode));
641 static inline bool ceph_dir_is_complete_ordered(struct inode *inode)
643 bool ret = __ceph_dir_is_complete_ordered(ceph_inode(inode));
648 /* find a specific frag @f */
649 extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci,
653 * choose fragment for value @v. copy frag content to pfrag, if leaf
656 extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v,
657 struct ceph_inode_frag *pfrag,
660 static inline struct ceph_dentry_info *ceph_dentry(const struct dentry *dentry)
662 return (struct ceph_dentry_info *)dentry->d_fsdata;
668 static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci)
670 return !RB_EMPTY_ROOT(&ci->i_caps);
673 extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented);
674 extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t);
675 extern int __ceph_caps_issued_mask_metric(struct ceph_inode_info *ci, int mask,
677 extern int __ceph_caps_issued_other(struct ceph_inode_info *ci,
678 struct ceph_cap *cap);
680 static inline int ceph_caps_issued(struct ceph_inode_info *ci)
683 spin_lock(&ci->i_ceph_lock);
684 issued = __ceph_caps_issued(ci, NULL);
685 spin_unlock(&ci->i_ceph_lock);
689 static inline int ceph_caps_issued_mask_metric(struct ceph_inode_info *ci,
693 spin_lock(&ci->i_ceph_lock);
694 r = __ceph_caps_issued_mask_metric(ci, mask, touch);
695 spin_unlock(&ci->i_ceph_lock);
699 static inline int __ceph_caps_dirty(struct ceph_inode_info *ci)
701 return ci->i_dirty_caps | ci->i_flushing_caps;
703 extern struct ceph_cap_flush *ceph_alloc_cap_flush(void);
704 extern void ceph_free_cap_flush(struct ceph_cap_flush *cf);
705 extern int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask,
706 struct ceph_cap_flush **pcf);
708 extern int __ceph_caps_revoking_other(struct ceph_inode_info *ci,
709 struct ceph_cap *ocap, int mask);
710 extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask);
711 extern int __ceph_caps_used(struct ceph_inode_info *ci);
713 static inline bool __ceph_is_file_opened(struct ceph_inode_info *ci)
715 return ci->i_nr_by_mode[0];
717 extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci);
718 extern int __ceph_caps_wanted(struct ceph_inode_info *ci);
720 /* what the mds thinks we want */
721 extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci, bool check);
723 extern void ceph_caps_init(struct ceph_mds_client *mdsc);
724 extern void ceph_caps_finalize(struct ceph_mds_client *mdsc);
725 extern void ceph_adjust_caps_max_min(struct ceph_mds_client *mdsc,
726 struct ceph_mount_options *fsopt);
727 extern int ceph_reserve_caps(struct ceph_mds_client *mdsc,
728 struct ceph_cap_reservation *ctx, int need);
729 extern void ceph_unreserve_caps(struct ceph_mds_client *mdsc,
730 struct ceph_cap_reservation *ctx);
731 extern void ceph_reservation_status(struct ceph_fs_client *client,
732 int *total, int *avail, int *used,
733 int *reserved, int *min);
738 * we keep buffered readdir results attached to file->private_data
740 #define CEPH_F_SYNC 1
741 #define CEPH_F_ATEND 2
743 struct ceph_file_info {
744 short fmode; /* initialized on open */
745 short flags; /* CEPH_F_* */
747 spinlock_t rw_contexts_lock;
748 struct list_head rw_contexts;
755 struct ceph_dir_file_info {
756 struct ceph_file_info file_info;
758 /* readdir: position within the dir */
760 struct ceph_mds_request *last_readdir;
762 /* readdir: position within a frag */
763 unsigned next_offset; /* offset of next chunk (last_name's + 1) */
764 char *last_name; /* last entry in previous chunk */
765 long long dir_release_count;
766 long long dir_ordered_count;
767 int readdir_cache_idx;
769 /* used for -o dirstat read() on directory thing */
774 struct ceph_rw_context {
775 struct list_head list;
776 struct task_struct *thread;
780 #define CEPH_DEFINE_RW_CONTEXT(_name, _caps) \
781 struct ceph_rw_context _name = { \
786 static inline void ceph_add_rw_context(struct ceph_file_info *cf,
787 struct ceph_rw_context *ctx)
789 spin_lock(&cf->rw_contexts_lock);
790 list_add(&ctx->list, &cf->rw_contexts);
791 spin_unlock(&cf->rw_contexts_lock);
794 static inline void ceph_del_rw_context(struct ceph_file_info *cf,
795 struct ceph_rw_context *ctx)
797 spin_lock(&cf->rw_contexts_lock);
798 list_del(&ctx->list);
799 spin_unlock(&cf->rw_contexts_lock);
802 static inline struct ceph_rw_context*
803 ceph_find_rw_context(struct ceph_file_info *cf)
805 struct ceph_rw_context *ctx, *found = NULL;
806 spin_lock(&cf->rw_contexts_lock);
807 list_for_each_entry(ctx, &cf->rw_contexts, list) {
808 if (ctx->thread == current) {
813 spin_unlock(&cf->rw_contexts_lock);
817 struct ceph_readdir_cache_control {
819 struct dentry **dentries;
824 * A "snap realm" describes a subset of the file hierarchy sharing
825 * the same set of snapshots that apply to it. The realms themselves
826 * are organized into a hierarchy, such that children inherit (some of)
827 * the snapshots of their parents.
829 * All inodes within the realm that have capabilities are linked into a
832 struct ceph_snap_realm {
840 u64 parent_since; /* snapid when our current parent became so */
842 u64 *prior_parent_snaps; /* snaps inherited from any parents we */
843 u32 num_prior_parent_snaps; /* had prior to parent_since */
844 u64 *snaps; /* snaps specific to this realm */
847 struct ceph_snap_realm *parent;
848 struct list_head children; /* list of child realms */
849 struct list_head child_item;
851 struct list_head empty_item; /* if i have ref==0 */
853 struct list_head dirty_item; /* if realm needs new context */
855 /* the current set of snaps for this realm */
856 struct ceph_snap_context *cached_context;
858 struct list_head inodes_with_caps;
859 spinlock_t inodes_with_caps_lock;
862 static inline int default_congestion_kb(void)
869 * congestion size, scale with available memory.
881 * This allows larger machines to have larger/more transfers.
882 * Limit the default to 256M
884 congestion_kb = (16*int_sqrt(totalram_pages())) << (PAGE_SHIFT-10);
885 if (congestion_kb > 256*1024)
886 congestion_kb = 256*1024;
888 return congestion_kb;
893 extern int ceph_force_reconnect(struct super_block *sb);
895 struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc,
897 extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc,
898 struct ceph_snap_realm *realm);
899 extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc,
900 struct ceph_snap_realm *realm);
901 extern int ceph_update_snap_trace(struct ceph_mds_client *m,
902 void *p, void *e, bool deletion,
903 struct ceph_snap_realm **realm_ret);
904 extern void ceph_handle_snap(struct ceph_mds_client *mdsc,
905 struct ceph_mds_session *session,
906 struct ceph_msg *msg);
907 extern void ceph_queue_cap_snap(struct ceph_inode_info *ci);
908 extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci,
909 struct ceph_cap_snap *capsnap);
910 extern void ceph_cleanup_empty_realms(struct ceph_mds_client *mdsc);
912 extern struct ceph_snapid_map *ceph_get_snapid_map(struct ceph_mds_client *mdsc,
914 extern void ceph_put_snapid_map(struct ceph_mds_client* mdsc,
915 struct ceph_snapid_map *sm);
916 extern void ceph_trim_snapid_map(struct ceph_mds_client *mdsc);
917 extern void ceph_cleanup_snapid_map(struct ceph_mds_client *mdsc);
921 * a cap_snap is "pending" if it is still awaiting an in-progress
922 * sync write (that may/may not still update size, mtime, etc.).
924 static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci)
926 return !list_empty(&ci->i_cap_snaps) &&
927 list_last_entry(&ci->i_cap_snaps, struct ceph_cap_snap,
932 struct ceph_mds_reply_info_in;
933 struct ceph_mds_reply_dirfrag;
935 extern const struct inode_operations ceph_file_iops;
937 extern struct inode *ceph_alloc_inode(struct super_block *sb);
938 extern void ceph_evict_inode(struct inode *inode);
939 extern void ceph_free_inode(struct inode *inode);
941 extern struct inode *ceph_get_inode(struct super_block *sb,
942 struct ceph_vino vino);
943 extern struct inode *ceph_get_snapdir(struct inode *parent);
944 extern int ceph_fill_file_size(struct inode *inode, int issued,
945 u32 truncate_seq, u64 truncate_size, u64 size);
946 extern void ceph_fill_file_time(struct inode *inode, int issued,
947 u64 time_warp_seq, struct timespec64 *ctime,
948 struct timespec64 *mtime,
949 struct timespec64 *atime);
950 extern int ceph_fill_inode(struct inode *inode, struct page *locked_page,
951 struct ceph_mds_reply_info_in *iinfo,
952 struct ceph_mds_reply_dirfrag *dirinfo,
953 struct ceph_mds_session *session, int cap_fmode,
954 struct ceph_cap_reservation *caps_reservation);
955 extern int ceph_fill_trace(struct super_block *sb,
956 struct ceph_mds_request *req);
957 extern int ceph_readdir_prepopulate(struct ceph_mds_request *req,
958 struct ceph_mds_session *session);
960 extern int ceph_inode_holds_cap(struct inode *inode, int mask);
962 extern bool ceph_inode_set_size(struct inode *inode, loff_t size);
963 extern void __ceph_do_pending_vmtruncate(struct inode *inode);
964 extern void ceph_queue_vmtruncate(struct inode *inode);
965 extern void ceph_queue_invalidate(struct inode *inode);
966 extern void ceph_queue_writeback(struct inode *inode);
967 extern void ceph_async_iput(struct inode *inode);
969 extern int __ceph_do_getattr(struct inode *inode, struct page *locked_page,
970 int mask, bool force);
971 static inline int ceph_do_getattr(struct inode *inode, int mask, bool force)
973 return __ceph_do_getattr(inode, NULL, mask, force);
975 extern int ceph_permission(struct inode *inode, int mask);
976 extern int __ceph_setattr(struct inode *inode, struct iattr *attr);
977 extern int ceph_setattr(struct dentry *dentry, struct iattr *attr);
978 extern int ceph_getattr(const struct path *path, struct kstat *stat,
979 u32 request_mask, unsigned int flags);
982 int __ceph_setxattr(struct inode *, const char *, const void *, size_t, int);
983 ssize_t __ceph_getxattr(struct inode *, const char *, void *, size_t);
984 extern ssize_t ceph_listxattr(struct dentry *, char *, size_t);
985 extern struct ceph_buffer *__ceph_build_xattrs_blob(struct ceph_inode_info *ci);
986 extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci);
987 extern const struct xattr_handler *ceph_xattr_handlers[];
989 struct ceph_acl_sec_ctx {
990 #ifdef CONFIG_CEPH_FS_POSIX_ACL
994 #ifdef CONFIG_CEPH_FS_SECURITY_LABEL
998 struct ceph_pagelist *pagelist;
1001 #ifdef CONFIG_SECURITY
1002 extern bool ceph_security_xattr_deadlock(struct inode *in);
1003 extern bool ceph_security_xattr_wanted(struct inode *in);
1005 static inline bool ceph_security_xattr_deadlock(struct inode *in)
1009 static inline bool ceph_security_xattr_wanted(struct inode *in)
1015 #ifdef CONFIG_CEPH_FS_SECURITY_LABEL
1016 extern int ceph_security_init_secctx(struct dentry *dentry, umode_t mode,
1017 struct ceph_acl_sec_ctx *ctx);
1018 static inline void ceph_security_invalidate_secctx(struct inode *inode)
1020 security_inode_invalidate_secctx(inode);
1023 static inline int ceph_security_init_secctx(struct dentry *dentry, umode_t mode,
1024 struct ceph_acl_sec_ctx *ctx)
1028 static inline void ceph_security_invalidate_secctx(struct inode *inode)
1033 void ceph_release_acl_sec_ctx(struct ceph_acl_sec_ctx *as_ctx);
1036 #ifdef CONFIG_CEPH_FS_POSIX_ACL
1038 struct posix_acl *ceph_get_acl(struct inode *, int);
1039 int ceph_set_acl(struct inode *inode, struct posix_acl *acl, int type);
1040 int ceph_pre_init_acls(struct inode *dir, umode_t *mode,
1041 struct ceph_acl_sec_ctx *as_ctx);
1042 void ceph_init_inode_acls(struct inode *inode,
1043 struct ceph_acl_sec_ctx *as_ctx);
1045 static inline void ceph_forget_all_cached_acls(struct inode *inode)
1047 forget_all_cached_acls(inode);
1052 #define ceph_get_acl NULL
1053 #define ceph_set_acl NULL
1055 static inline int ceph_pre_init_acls(struct inode *dir, umode_t *mode,
1056 struct ceph_acl_sec_ctx *as_ctx)
1060 static inline void ceph_init_inode_acls(struct inode *inode,
1061 struct ceph_acl_sec_ctx *as_ctx)
1064 static inline int ceph_acl_chmod(struct dentry *dentry, struct inode *inode)
1069 static inline void ceph_forget_all_cached_acls(struct inode *inode)
1076 extern const char *ceph_cap_string(int c);
1077 extern void ceph_handle_caps(struct ceph_mds_session *session,
1078 struct ceph_msg *msg);
1079 extern struct ceph_cap *ceph_get_cap(struct ceph_mds_client *mdsc,
1080 struct ceph_cap_reservation *ctx);
1081 extern void ceph_add_cap(struct inode *inode,
1082 struct ceph_mds_session *session, u64 cap_id,
1083 unsigned issued, unsigned wanted,
1084 unsigned cap, unsigned seq, u64 realmino, int flags,
1085 struct ceph_cap **new_cap);
1086 extern void __ceph_remove_cap(struct ceph_cap *cap, bool queue_release);
1087 extern void __ceph_remove_caps(struct ceph_inode_info *ci);
1088 extern void ceph_put_cap(struct ceph_mds_client *mdsc,
1089 struct ceph_cap *cap);
1090 extern int ceph_is_any_caps(struct inode *inode);
1092 extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc);
1093 extern int ceph_fsync(struct file *file, loff_t start, loff_t end,
1095 extern void ceph_early_kick_flushing_caps(struct ceph_mds_client *mdsc,
1096 struct ceph_mds_session *session);
1097 extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc,
1098 struct ceph_mds_session *session);
1099 void ceph_kick_flushing_inode_caps(struct ceph_mds_session *session,
1100 struct ceph_inode_info *ci);
1101 extern struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci,
1103 extern void ceph_take_cap_refs(struct ceph_inode_info *ci, int caps,
1104 bool snap_rwsem_locked);
1105 extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps);
1106 extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had);
1107 extern void ceph_put_cap_refs_no_check_caps(struct ceph_inode_info *ci,
1109 extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
1110 struct ceph_snap_context *snapc);
1111 extern void ceph_flush_snaps(struct ceph_inode_info *ci,
1112 struct ceph_mds_session **psession);
1113 extern bool __ceph_should_report_size(struct ceph_inode_info *ci);
1114 extern void ceph_check_caps(struct ceph_inode_info *ci, int flags,
1115 struct ceph_mds_session *session);
1116 extern void ceph_check_delayed_caps(struct ceph_mds_client *mdsc);
1117 extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc);
1118 extern int ceph_drop_caps_for_unlink(struct inode *inode);
1119 extern int ceph_encode_inode_release(void **p, struct inode *inode,
1120 int mds, int drop, int unless, int force);
1121 extern int ceph_encode_dentry_release(void **p, struct dentry *dn,
1123 int mds, int drop, int unless);
1125 extern int ceph_get_caps(struct file *filp, int need, int want,
1126 loff_t endoff, int *got, struct page **pinned_page);
1127 extern int ceph_try_get_caps(struct inode *inode,
1128 int need, int want, bool nonblock, int *got);
1130 /* for counting open files by mode */
1131 extern void ceph_get_fmode(struct ceph_inode_info *ci, int mode, int count);
1132 extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode, int count);
1133 extern void __ceph_touch_fmode(struct ceph_inode_info *ci,
1134 struct ceph_mds_client *mdsc, int fmode);
1137 extern const struct address_space_operations ceph_aops;
1138 extern int ceph_mmap(struct file *file, struct vm_area_struct *vma);
1139 extern int ceph_uninline_data(struct file *filp, struct page *locked_page);
1140 extern int ceph_pool_perm_check(struct inode *inode, int need);
1141 extern void ceph_pool_perm_destroy(struct ceph_mds_client* mdsc);
1144 extern const struct file_operations ceph_file_fops;
1146 extern int ceph_renew_caps(struct inode *inode, int fmode);
1147 extern int ceph_open(struct inode *inode, struct file *file);
1148 extern int ceph_atomic_open(struct inode *dir, struct dentry *dentry,
1149 struct file *file, unsigned flags, umode_t mode);
1150 extern int ceph_release(struct inode *inode, struct file *filp);
1151 extern void ceph_fill_inline_data(struct inode *inode, struct page *locked_page,
1152 char *data, size_t len);
1155 extern const struct file_operations ceph_dir_fops;
1156 extern const struct file_operations ceph_snapdir_fops;
1157 extern const struct inode_operations ceph_dir_iops;
1158 extern const struct inode_operations ceph_snapdir_iops;
1159 extern const struct dentry_operations ceph_dentry_ops;
1161 extern loff_t ceph_make_fpos(unsigned high, unsigned off, bool hash_order);
1162 extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry);
1163 extern int ceph_handle_snapdir(struct ceph_mds_request *req,
1164 struct dentry *dentry, int err);
1165 extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req,
1166 struct dentry *dentry, int err);
1168 extern void __ceph_dentry_lease_touch(struct ceph_dentry_info *di);
1169 extern void __ceph_dentry_dir_lease_touch(struct ceph_dentry_info *di);
1170 extern void ceph_invalidate_dentry_lease(struct dentry *dentry);
1171 extern int ceph_trim_dentries(struct ceph_mds_client *mdsc);
1172 extern unsigned ceph_dentry_hash(struct inode *dir, struct dentry *dn);
1173 extern void ceph_readdir_cache_release(struct ceph_readdir_cache_control *ctl);
1176 extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
1179 extern const struct export_operations ceph_export_ops;
1180 struct inode *ceph_lookup_inode(struct super_block *sb, u64 ino);
1183 extern __init void ceph_flock_init(void);
1184 extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl);
1185 extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl);
1186 extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num);
1187 extern int ceph_encode_locks_to_buffer(struct inode *inode,
1188 struct ceph_filelock *flocks,
1189 int num_fcntl_locks,
1190 int num_flock_locks);
1191 extern int ceph_locks_to_pagelist(struct ceph_filelock *flocks,
1192 struct ceph_pagelist *pagelist,
1193 int num_fcntl_locks, int num_flock_locks);
1196 extern void ceph_fs_debugfs_init(struct ceph_fs_client *client);
1197 extern void ceph_fs_debugfs_cleanup(struct ceph_fs_client *client);
1200 static inline bool __ceph_has_any_quota(struct ceph_inode_info *ci)
1202 return ci->i_max_files || ci->i_max_bytes;
1205 extern void ceph_adjust_quota_realms_count(struct inode *inode, bool inc);
1207 static inline void __ceph_update_quota(struct ceph_inode_info *ci,
1208 u64 max_bytes, u64 max_files)
1210 bool had_quota, has_quota;
1211 had_quota = __ceph_has_any_quota(ci);
1212 ci->i_max_bytes = max_bytes;
1213 ci->i_max_files = max_files;
1214 has_quota = __ceph_has_any_quota(ci);
1216 if (had_quota != has_quota)
1217 ceph_adjust_quota_realms_count(&ci->vfs_inode, has_quota);
1220 extern void ceph_handle_quota(struct ceph_mds_client *mdsc,
1221 struct ceph_mds_session *session,
1222 struct ceph_msg *msg);
1223 extern bool ceph_quota_is_max_files_exceeded(struct inode *inode);
1224 extern bool ceph_quota_is_max_bytes_exceeded(struct inode *inode,
1226 extern bool ceph_quota_is_max_bytes_approaching(struct inode *inode,
1228 extern bool ceph_quota_update_statfs(struct ceph_fs_client *fsc,
1229 struct kstatfs *buf);
1230 extern int ceph_quota_check_rename(struct ceph_mds_client *mdsc,
1231 struct inode *old, struct inode *new);
1232 extern void ceph_cleanup_quotarealms_inodes(struct ceph_mds_client *mdsc);
1234 #endif /* _FS_CEPH_SUPER_H */