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>
20 #include <linux/ceph/libceph.h>
22 #ifdef CONFIG_CEPH_FSCACHE
23 #include <linux/fscache.h>
26 /* f_type in struct statfs */
27 #define CEPH_SUPER_MAGIC 0x00c36400
29 /* large granularity for statfs utilization stats to facilitate
30 * large volume sizes on 32-bit machines. */
31 #define CEPH_BLOCK_SHIFT 22 /* 4 MB */
32 #define CEPH_BLOCK (1 << CEPH_BLOCK_SHIFT)
34 #define CEPH_MOUNT_OPT_DIRSTAT (1<<4) /* `cat dirname` for stats */
35 #define CEPH_MOUNT_OPT_RBYTES (1<<5) /* dir st_bytes = rbytes */
36 #define CEPH_MOUNT_OPT_NOASYNCREADDIR (1<<7) /* no dcache readdir */
37 #define CEPH_MOUNT_OPT_INO32 (1<<8) /* 32 bit inos */
38 #define CEPH_MOUNT_OPT_DCACHE (1<<9) /* use dcache for readdir etc */
39 #define CEPH_MOUNT_OPT_FSCACHE (1<<10) /* use fscache */
40 #define CEPH_MOUNT_OPT_NOPOOLPERM (1<<11) /* no pool permission check */
41 #define CEPH_MOUNT_OPT_MOUNTWAIT (1<<12) /* mount waits if no mds is up */
42 #define CEPH_MOUNT_OPT_NOQUOTADF (1<<13) /* no root dir quota in statfs */
43 #define CEPH_MOUNT_OPT_NOCOPYFROM (1<<14) /* don't use RADOS 'copy-from' op */
45 #define CEPH_MOUNT_OPT_DEFAULT \
46 (CEPH_MOUNT_OPT_DCACHE | \
47 CEPH_MOUNT_OPT_NOCOPYFROM)
49 #define ceph_set_mount_opt(fsc, opt) \
50 (fsc)->mount_options->flags |= CEPH_MOUNT_OPT_##opt;
51 #define ceph_test_mount_opt(fsc, opt) \
52 (!!((fsc)->mount_options->flags & CEPH_MOUNT_OPT_##opt))
54 /* max size of osd read request, limited by libceph */
55 #define CEPH_MAX_READ_SIZE CEPH_MSG_MAX_DATA_LEN
56 /* osd has a configurable limitaion of max write size.
57 * CEPH_MSG_MAX_DATA_LEN should be small enough. */
58 #define CEPH_MAX_WRITE_SIZE CEPH_MSG_MAX_DATA_LEN
59 #define CEPH_RASIZE_DEFAULT (8192*1024) /* max readahead */
60 #define CEPH_MAX_READDIR_DEFAULT 1024
61 #define CEPH_MAX_READDIR_BYTES_DEFAULT (512*1024)
62 #define CEPH_SNAPDIRNAME_DEFAULT ".snap"
65 * Delay telling the MDS we no longer want caps, in case we reopen
66 * the file. Delay a minimum amount of time, even if we send a cap
67 * message for some other reason. Otherwise, take the oppotunity to
68 * update the mds to avoid sending another message later.
70 #define CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT 5 /* cap release delay */
71 #define CEPH_CAPS_WANTED_DELAY_MAX_DEFAULT 60 /* cap release delay */
73 struct ceph_mount_options {
77 int wsize; /* max write size */
78 int rsize; /* max read size */
79 int rasize; /* max readahead */
80 int congestion_kb; /* max writeback in flight */
81 int caps_wanted_delay_min, caps_wanted_delay_max;
82 int max_readdir; /* max readdir result (entires) */
83 int max_readdir_bytes; /* max readdir result (bytes) */
86 * everything above this point can be memcmp'd; everything below
87 * is handled in compare_mount_options()
90 char *snapdir_name; /* default ".snap" */
91 char *mds_namespace; /* default NULL */
92 char *server_path; /* default "/" */
93 char *fscache_uniq; /* default NULL */
96 struct ceph_fs_client {
97 struct super_block *sb;
99 struct ceph_mount_options *mount_options;
100 struct ceph_client *client;
102 unsigned long mount_state;
103 int min_caps; /* min caps i added */
104 loff_t max_file_size;
106 struct ceph_mds_client *mdsc;
109 mempool_t *wb_pagevec_pool;
110 struct workqueue_struct *wb_wq;
111 struct workqueue_struct *pg_inv_wq;
112 struct workqueue_struct *trunc_wq;
113 atomic_long_t writeback_count;
115 #ifdef CONFIG_DEBUG_FS
116 struct dentry *debugfs_dentry_lru, *debugfs_caps;
117 struct dentry *debugfs_congestion_kb;
118 struct dentry *debugfs_bdi;
119 struct dentry *debugfs_mdsc, *debugfs_mdsmap;
120 struct dentry *debugfs_mds_sessions;
123 #ifdef CONFIG_CEPH_FSCACHE
124 struct fscache_cookie *fscache;
130 * File i/o capability. This tracks shared state with the metadata
131 * server that allows us to cache or writeback attributes or to read
132 * and write data. For any given inode, we should have one or more
133 * capabilities, one issued by each metadata server, and our
134 * cumulative access is the OR of all issued capabilities.
136 * Each cap is referenced by the inode's i_caps rbtree and by per-mds
137 * session capability lists.
140 struct ceph_inode_info *ci;
141 struct rb_node ci_node; /* per-ci cap tree */
142 struct ceph_mds_session *session;
143 struct list_head session_caps; /* per-session caplist */
144 u64 cap_id; /* unique cap id (mds provided) */
148 int issued; /* latest, from the mds */
149 int implemented; /* implemented superset of
150 issued (for revocation) */
153 /* caps to release */
159 u32 seq, issue_seq, mseq;
160 u32 cap_gen; /* active/stale cycle */
161 unsigned long last_used;
162 struct list_head caps_item;
165 #define CHECK_CAPS_NODELAY 1 /* do not delay any further */
166 #define CHECK_CAPS_AUTHONLY 2 /* only check auth cap */
167 #define CHECK_CAPS_FLUSH 4 /* flush any dirty caps */
169 struct ceph_cap_flush {
171 int caps; /* 0 means capsnap */
172 bool wake; /* wake up flush waiters when finish ? */
173 struct list_head g_list; // global
174 struct list_head i_list; // per inode
178 * Snapped cap state that is pending flush to mds. When a snapshot occurs,
179 * we first complete any in-process sync writes and writeback any dirty
180 * data before flushing the snapped state (tracked here) back to the MDS.
182 struct ceph_cap_snap {
184 struct list_head ci_item;
186 struct ceph_cap_flush cap_flush;
190 struct ceph_snap_context *context;
196 struct ceph_buffer *xattr_blob;
200 struct timespec64 mtime, atime, ctime;
204 int writing; /* a sync write is still in progress */
205 int dirty_pages; /* dirty pages awaiting writeback */
210 static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap)
212 if (refcount_dec_and_test(&capsnap->nref)) {
213 if (capsnap->xattr_blob)
214 ceph_buffer_put(capsnap->xattr_blob);
220 * The frag tree describes how a directory is fragmented, potentially across
221 * multiple metadata servers. It is also used to indicate points where
222 * metadata authority is delegated, and whether/where metadata is replicated.
224 * A _leaf_ frag will be present in the i_fragtree IFF there is
225 * delegation info. That is, if mds >= 0 || ndist > 0.
227 #define CEPH_MAX_DIRFRAG_REP 4
229 struct ceph_inode_frag {
234 int split_by; /* i.e. 2^(split_by) children */
236 /* delegation and replication info */
237 int mds; /* -1 if same authority as parent */
238 int ndist; /* >0 if replicated */
239 int dist[CEPH_MAX_DIRFRAG_REP];
243 * We cache inode xattrs as an encoded blob until they are first used,
244 * at which point we parse them into an rbtree.
246 struct ceph_inode_xattr {
255 int should_free_name;
262 struct ceph_dentry_info {
263 struct ceph_mds_session *lease_session;
264 int lease_shared_gen;
267 unsigned long lease_renew_after, lease_renew_from;
268 struct list_head lru;
269 struct dentry *dentry;
274 struct ceph_inode_xattrs_info {
276 * (still encoded) xattr blob. we avoid the overhead of parsing
277 * this until someone actually calls getxattr, etc.
279 * blob->vec.iov_len == 4 implies there are no xattrs; blob ==
280 * NULL means we don't know.
282 struct ceph_buffer *blob, *prealloc_blob;
284 struct rb_root index;
289 u64 version, index_version;
295 struct ceph_inode_info {
296 struct ceph_vino i_vino; /* ceph ino + snap */
298 spinlock_t i_ceph_lock;
301 u64 i_inline_version;
304 unsigned i_ceph_flags;
305 atomic64_t i_release_count;
306 atomic64_t i_ordered_count;
307 atomic64_t i_complete_seq[2];
309 struct ceph_dir_layout i_dir_layout;
310 struct ceph_file_layout i_layout;
314 struct timespec64 i_rctime;
315 u64 i_rbytes, i_rfiles, i_rsubdirs;
316 u64 i_files, i_subdirs;
319 u64 i_max_bytes, i_max_files;
321 struct rb_root i_fragtree;
322 int i_fragtree_nsplits;
323 struct mutex i_fragtree_mutex;
325 struct ceph_inode_xattrs_info i_xattrs;
327 /* capabilities. protected _both_ by i_ceph_lock and cap->session's
329 struct rb_root i_caps; /* cap list */
330 struct ceph_cap *i_auth_cap; /* authoritative cap, if any */
331 unsigned i_dirty_caps, i_flushing_caps; /* mask of dirtied fields */
332 struct list_head i_dirty_item, i_flushing_item;
333 /* we need to track cap writeback on a per-cap-bit basis, to allow
334 * overlapping, pipelined cap flushes to the mds. we can probably
335 * reduce the tid to 8 bits if we're concerned about inode size. */
336 struct ceph_cap_flush *i_prealloc_cap_flush;
337 struct list_head i_cap_flush_list;
338 wait_queue_head_t i_cap_wq; /* threads waiting on a capability */
339 unsigned long i_hold_caps_min; /* jiffies */
340 unsigned long i_hold_caps_max; /* jiffies */
341 struct list_head i_cap_delay_list; /* for delayed cap release to mds */
342 struct ceph_cap_reservation i_cap_migration_resv;
343 struct list_head i_cap_snaps; /* snapped state pending flush to mds */
344 struct ceph_snap_context *i_head_snapc; /* set if wr_buffer_head > 0 or
345 dirty|flushing caps */
346 unsigned i_snap_caps; /* cap bits for snapped files */
348 int i_nr_by_mode[CEPH_FILE_MODE_BITS]; /* open file counts */
350 struct mutex i_truncate_mutex;
351 u32 i_truncate_seq; /* last truncate to smaller size */
352 u64 i_truncate_size; /* and the size we last truncated down to */
353 int i_truncate_pending; /* still need to call vmtruncate */
355 u64 i_max_size; /* max file size authorized by mds */
356 u64 i_reported_size; /* (max_)size reported to or requested of mds */
357 u64 i_wanted_max_size; /* offset we'd like to write too */
358 u64 i_requested_max_size; /* max_size we've requested */
360 /* held references to caps */
362 int i_rd_ref, i_rdcache_ref, i_wr_ref, i_wb_ref;
363 int i_wrbuffer_ref, i_wrbuffer_ref_head;
364 atomic_t i_filelock_ref;
365 atomic_t i_shared_gen; /* increment each time we get FILE_SHARED */
366 u32 i_rdcache_gen; /* incremented each time we get FILE_CACHE. */
367 u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */
369 struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */
370 struct list_head i_unsafe_iops; /* uncommitted mds inode ops */
371 spinlock_t i_unsafe_lock;
373 struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */
374 int i_snap_realm_counter; /* snap realm (if caps) */
375 struct list_head i_snap_realm_item;
376 struct list_head i_snap_flush_item;
378 struct work_struct i_wb_work; /* writeback work */
379 struct work_struct i_pg_inv_work; /* page invalidation work */
381 struct work_struct i_vmtruncate_work;
383 #ifdef CONFIG_CEPH_FSCACHE
384 struct fscache_cookie *fscache;
387 struct inode vfs_inode; /* at end */
390 static inline struct ceph_inode_info *ceph_inode(struct inode *inode)
392 return container_of(inode, struct ceph_inode_info, vfs_inode);
395 static inline struct ceph_fs_client *ceph_inode_to_client(struct inode *inode)
397 return (struct ceph_fs_client *)inode->i_sb->s_fs_info;
400 static inline struct ceph_fs_client *ceph_sb_to_client(struct super_block *sb)
402 return (struct ceph_fs_client *)sb->s_fs_info;
405 static inline struct ceph_vino ceph_vino(struct inode *inode)
407 return ceph_inode(inode)->i_vino;
411 * ino_t is <64 bits on many architectures, blech.
413 * i_ino (kernel inode) st_ino (userspace)
418 static inline u32 ceph_ino_to_ino32(__u64 vino)
420 u32 ino = vino & 0xffffffff;
430 static inline ino_t ceph_vino_to_ino(struct ceph_vino vino)
432 #if BITS_PER_LONG == 32
433 return ceph_ino_to_ino32(vino.ino);
435 return (ino_t)vino.ino;
440 * user-visible ino (stat, filldir)
442 #if BITS_PER_LONG == 32
443 static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino)
448 static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino)
450 if (ceph_test_mount_opt(ceph_sb_to_client(sb), INO32))
451 ino = ceph_ino_to_ino32(ino);
457 /* for printf-style formatting */
458 #define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap
460 static inline u64 ceph_ino(struct inode *inode)
462 return ceph_inode(inode)->i_vino.ino;
464 static inline u64 ceph_snap(struct inode *inode)
466 return ceph_inode(inode)->i_vino.snap;
469 static inline int ceph_ino_compare(struct inode *inode, void *data)
471 struct ceph_vino *pvino = (struct ceph_vino *)data;
472 struct ceph_inode_info *ci = ceph_inode(inode);
473 return ci->i_vino.ino == pvino->ino &&
474 ci->i_vino.snap == pvino->snap;
477 static inline struct inode *ceph_find_inode(struct super_block *sb,
478 struct ceph_vino vino)
480 ino_t t = ceph_vino_to_ino(vino);
481 return ilookup5(sb, t, ceph_ino_compare, &vino);
488 #define CEPH_I_DIR_ORDERED (1 << 0) /* dentries in dir are ordered */
489 #define CEPH_I_NODELAY (1 << 1) /* do not delay cap release */
490 #define CEPH_I_FLUSH (1 << 2) /* do not delay flush of dirty metadata */
491 #define CEPH_I_NOFLUSH (1 << 3) /* do not flush dirty caps */
492 #define CEPH_I_POOL_PERM (1 << 4) /* pool rd/wr bits are valid */
493 #define CEPH_I_POOL_RD (1 << 5) /* can read from pool */
494 #define CEPH_I_POOL_WR (1 << 6) /* can write to pool */
495 #define CEPH_I_SEC_INITED (1 << 7) /* security initialized */
496 #define CEPH_I_CAP_DROPPED (1 << 8) /* caps were forcibly dropped */
497 #define CEPH_I_KICK_FLUSH (1 << 9) /* kick flushing caps */
498 #define CEPH_I_FLUSH_SNAPS (1 << 10) /* need flush snapss */
499 #define CEPH_I_ERROR_WRITE (1 << 11) /* have seen write errors */
500 #define CEPH_I_ERROR_FILELOCK (1 << 12) /* have seen file lock errors */
504 * We set the ERROR_WRITE bit when we start seeing write errors on an inode
505 * and then clear it when they start succeeding. Note that we do a lockless
506 * check first, and only take the lock if it looks like it needs to be changed.
507 * The write submission code just takes this as a hint, so we're not too
508 * worried if a few slip through in either direction.
510 static inline void ceph_set_error_write(struct ceph_inode_info *ci)
512 if (!(READ_ONCE(ci->i_ceph_flags) & CEPH_I_ERROR_WRITE)) {
513 spin_lock(&ci->i_ceph_lock);
514 ci->i_ceph_flags |= CEPH_I_ERROR_WRITE;
515 spin_unlock(&ci->i_ceph_lock);
519 static inline void ceph_clear_error_write(struct ceph_inode_info *ci)
521 if (READ_ONCE(ci->i_ceph_flags) & CEPH_I_ERROR_WRITE) {
522 spin_lock(&ci->i_ceph_lock);
523 ci->i_ceph_flags &= ~CEPH_I_ERROR_WRITE;
524 spin_unlock(&ci->i_ceph_lock);
528 static inline void __ceph_dir_set_complete(struct ceph_inode_info *ci,
529 long long release_count,
530 long long ordered_count)
532 smp_mb__before_atomic();
533 atomic64_set(&ci->i_complete_seq[0], release_count);
534 atomic64_set(&ci->i_complete_seq[1], ordered_count);
537 static inline void __ceph_dir_clear_complete(struct ceph_inode_info *ci)
539 atomic64_inc(&ci->i_release_count);
542 static inline void __ceph_dir_clear_ordered(struct ceph_inode_info *ci)
544 atomic64_inc(&ci->i_ordered_count);
547 static inline bool __ceph_dir_is_complete(struct ceph_inode_info *ci)
549 return atomic64_read(&ci->i_complete_seq[0]) ==
550 atomic64_read(&ci->i_release_count);
553 static inline bool __ceph_dir_is_complete_ordered(struct ceph_inode_info *ci)
555 return atomic64_read(&ci->i_complete_seq[0]) ==
556 atomic64_read(&ci->i_release_count) &&
557 atomic64_read(&ci->i_complete_seq[1]) ==
558 atomic64_read(&ci->i_ordered_count);
561 static inline void ceph_dir_clear_complete(struct inode *inode)
563 __ceph_dir_clear_complete(ceph_inode(inode));
566 static inline void ceph_dir_clear_ordered(struct inode *inode)
568 __ceph_dir_clear_ordered(ceph_inode(inode));
571 static inline bool ceph_dir_is_complete_ordered(struct inode *inode)
573 bool ret = __ceph_dir_is_complete_ordered(ceph_inode(inode));
578 /* find a specific frag @f */
579 extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci,
583 * choose fragment for value @v. copy frag content to pfrag, if leaf
586 extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v,
587 struct ceph_inode_frag *pfrag,
590 static inline struct ceph_dentry_info *ceph_dentry(struct dentry *dentry)
592 return (struct ceph_dentry_info *)dentry->d_fsdata;
598 static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci)
600 return !RB_EMPTY_ROOT(&ci->i_caps);
603 extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented);
604 extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t);
605 extern int __ceph_caps_issued_other(struct ceph_inode_info *ci,
606 struct ceph_cap *cap);
608 static inline int ceph_caps_issued(struct ceph_inode_info *ci)
611 spin_lock(&ci->i_ceph_lock);
612 issued = __ceph_caps_issued(ci, NULL);
613 spin_unlock(&ci->i_ceph_lock);
617 static inline int ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask,
621 spin_lock(&ci->i_ceph_lock);
622 r = __ceph_caps_issued_mask(ci, mask, touch);
623 spin_unlock(&ci->i_ceph_lock);
627 static inline int __ceph_caps_dirty(struct ceph_inode_info *ci)
629 return ci->i_dirty_caps | ci->i_flushing_caps;
631 extern struct ceph_cap_flush *ceph_alloc_cap_flush(void);
632 extern void ceph_free_cap_flush(struct ceph_cap_flush *cf);
633 extern int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask,
634 struct ceph_cap_flush **pcf);
636 extern int __ceph_caps_revoking_other(struct ceph_inode_info *ci,
637 struct ceph_cap *ocap, int mask);
638 extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask);
639 extern int __ceph_caps_used(struct ceph_inode_info *ci);
641 extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci);
644 * wanted, by virtue of open file modes AND cap refs (buffered/cached data)
646 static inline int __ceph_caps_wanted(struct ceph_inode_info *ci)
648 int w = __ceph_caps_file_wanted(ci) | __ceph_caps_used(ci);
649 if (w & CEPH_CAP_FILE_BUFFER)
650 w |= CEPH_CAP_FILE_EXCL; /* we want EXCL if dirty data */
654 /* what the mds thinks we want */
655 extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci, bool check);
657 extern void ceph_caps_init(struct ceph_mds_client *mdsc);
658 extern void ceph_caps_finalize(struct ceph_mds_client *mdsc);
659 extern void ceph_adjust_min_caps(struct ceph_mds_client *mdsc, int delta);
660 extern int ceph_reserve_caps(struct ceph_mds_client *mdsc,
661 struct ceph_cap_reservation *ctx, int need);
662 extern void ceph_unreserve_caps(struct ceph_mds_client *mdsc,
663 struct ceph_cap_reservation *ctx);
664 extern void ceph_reservation_status(struct ceph_fs_client *client,
665 int *total, int *avail, int *used,
666 int *reserved, int *min);
671 * we keep buffered readdir results attached to file->private_data
673 #define CEPH_F_SYNC 1
674 #define CEPH_F_ATEND 2
676 struct ceph_file_info {
677 short fmode; /* initialized on open */
678 short flags; /* CEPH_F_* */
680 spinlock_t rw_contexts_lock;
681 struct list_head rw_contexts;
684 struct ceph_dir_file_info {
685 struct ceph_file_info file_info;
687 /* readdir: position within the dir */
689 struct ceph_mds_request *last_readdir;
691 /* readdir: position within a frag */
692 unsigned next_offset; /* offset of next chunk (last_name's + 1) */
693 char *last_name; /* last entry in previous chunk */
694 long long dir_release_count;
695 long long dir_ordered_count;
696 int readdir_cache_idx;
698 /* used for -o dirstat read() on directory thing */
703 struct ceph_rw_context {
704 struct list_head list;
705 struct task_struct *thread;
709 #define CEPH_DEFINE_RW_CONTEXT(_name, _caps) \
710 struct ceph_rw_context _name = { \
715 static inline void ceph_add_rw_context(struct ceph_file_info *cf,
716 struct ceph_rw_context *ctx)
718 spin_lock(&cf->rw_contexts_lock);
719 list_add(&ctx->list, &cf->rw_contexts);
720 spin_unlock(&cf->rw_contexts_lock);
723 static inline void ceph_del_rw_context(struct ceph_file_info *cf,
724 struct ceph_rw_context *ctx)
726 spin_lock(&cf->rw_contexts_lock);
727 list_del(&ctx->list);
728 spin_unlock(&cf->rw_contexts_lock);
731 static inline struct ceph_rw_context*
732 ceph_find_rw_context(struct ceph_file_info *cf)
734 struct ceph_rw_context *ctx, *found = NULL;
735 spin_lock(&cf->rw_contexts_lock);
736 list_for_each_entry(ctx, &cf->rw_contexts, list) {
737 if (ctx->thread == current) {
742 spin_unlock(&cf->rw_contexts_lock);
746 struct ceph_readdir_cache_control {
748 struct dentry **dentries;
753 * A "snap realm" describes a subset of the file hierarchy sharing
754 * the same set of snapshots that apply to it. The realms themselves
755 * are organized into a hierarchy, such that children inherit (some of)
756 * the snapshots of their parents.
758 * All inodes within the realm that have capabilities are linked into a
761 struct ceph_snap_realm {
769 u64 parent_since; /* snapid when our current parent became so */
771 u64 *prior_parent_snaps; /* snaps inherited from any parents we */
772 u32 num_prior_parent_snaps; /* had prior to parent_since */
773 u64 *snaps; /* snaps specific to this realm */
776 struct ceph_snap_realm *parent;
777 struct list_head children; /* list of child realms */
778 struct list_head child_item;
780 struct list_head empty_item; /* if i have ref==0 */
782 struct list_head dirty_item; /* if realm needs new context */
784 /* the current set of snaps for this realm */
785 struct ceph_snap_context *cached_context;
787 struct list_head inodes_with_caps;
788 spinlock_t inodes_with_caps_lock;
791 static inline int default_congestion_kb(void)
798 * congestion size, scale with available memory.
810 * This allows larger machines to have larger/more transfers.
811 * Limit the default to 256M
813 congestion_kb = (16*int_sqrt(totalram_pages())) << (PAGE_SHIFT-10);
814 if (congestion_kb > 256*1024)
815 congestion_kb = 256*1024;
817 return congestion_kb;
823 struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc,
825 extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc,
826 struct ceph_snap_realm *realm);
827 extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc,
828 struct ceph_snap_realm *realm);
829 extern int ceph_update_snap_trace(struct ceph_mds_client *m,
830 void *p, void *e, bool deletion,
831 struct ceph_snap_realm **realm_ret);
832 extern void ceph_handle_snap(struct ceph_mds_client *mdsc,
833 struct ceph_mds_session *session,
834 struct ceph_msg *msg);
835 extern void ceph_queue_cap_snap(struct ceph_inode_info *ci);
836 extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci,
837 struct ceph_cap_snap *capsnap);
838 extern void ceph_cleanup_empty_realms(struct ceph_mds_client *mdsc);
841 * a cap_snap is "pending" if it is still awaiting an in-progress
842 * sync write (that may/may not still update size, mtime, etc.).
844 static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci)
846 return !list_empty(&ci->i_cap_snaps) &&
847 list_last_entry(&ci->i_cap_snaps, struct ceph_cap_snap,
852 extern const struct inode_operations ceph_file_iops;
854 extern struct inode *ceph_alloc_inode(struct super_block *sb);
855 extern void ceph_destroy_inode(struct inode *inode);
856 extern int ceph_drop_inode(struct inode *inode);
858 extern struct inode *ceph_get_inode(struct super_block *sb,
859 struct ceph_vino vino);
860 extern struct inode *ceph_get_snapdir(struct inode *parent);
861 extern int ceph_fill_file_size(struct inode *inode, int issued,
862 u32 truncate_seq, u64 truncate_size, u64 size);
863 extern void ceph_fill_file_time(struct inode *inode, int issued,
864 u64 time_warp_seq, struct timespec64 *ctime,
865 struct timespec64 *mtime,
866 struct timespec64 *atime);
867 extern int ceph_fill_trace(struct super_block *sb,
868 struct ceph_mds_request *req);
869 extern int ceph_readdir_prepopulate(struct ceph_mds_request *req,
870 struct ceph_mds_session *session);
872 extern int ceph_inode_holds_cap(struct inode *inode, int mask);
874 extern bool ceph_inode_set_size(struct inode *inode, loff_t size);
875 extern void __ceph_do_pending_vmtruncate(struct inode *inode);
876 extern void ceph_queue_vmtruncate(struct inode *inode);
878 extern void ceph_queue_invalidate(struct inode *inode);
879 extern void ceph_queue_writeback(struct inode *inode);
881 extern int __ceph_do_getattr(struct inode *inode, struct page *locked_page,
882 int mask, bool force);
883 static inline int ceph_do_getattr(struct inode *inode, int mask, bool force)
885 return __ceph_do_getattr(inode, NULL, mask, force);
887 extern int ceph_permission(struct inode *inode, int mask);
888 extern int __ceph_setattr(struct inode *inode, struct iattr *attr);
889 extern int ceph_setattr(struct dentry *dentry, struct iattr *attr);
890 extern int ceph_getattr(const struct path *path, struct kstat *stat,
891 u32 request_mask, unsigned int flags);
894 int __ceph_setxattr(struct inode *, const char *, const void *, size_t, int);
895 ssize_t __ceph_getxattr(struct inode *, const char *, void *, size_t);
896 extern ssize_t ceph_listxattr(struct dentry *, char *, size_t);
897 extern void __ceph_build_xattrs_blob(struct ceph_inode_info *ci);
898 extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci);
899 extern void __init ceph_xattr_init(void);
900 extern void ceph_xattr_exit(void);
901 extern const struct xattr_handler *ceph_xattr_handlers[];
903 #ifdef CONFIG_SECURITY
904 extern bool ceph_security_xattr_deadlock(struct inode *in);
905 extern bool ceph_security_xattr_wanted(struct inode *in);
907 static inline bool ceph_security_xattr_deadlock(struct inode *in)
911 static inline bool ceph_security_xattr_wanted(struct inode *in)
918 struct ceph_acls_info {
921 struct ceph_pagelist *pagelist;
924 #ifdef CONFIG_CEPH_FS_POSIX_ACL
926 struct posix_acl *ceph_get_acl(struct inode *, int);
927 int ceph_set_acl(struct inode *inode, struct posix_acl *acl, int type);
928 int ceph_pre_init_acls(struct inode *dir, umode_t *mode,
929 struct ceph_acls_info *info);
930 void ceph_init_inode_acls(struct inode *inode, struct ceph_acls_info *info);
931 void ceph_release_acls_info(struct ceph_acls_info *info);
933 static inline void ceph_forget_all_cached_acls(struct inode *inode)
935 forget_all_cached_acls(inode);
940 #define ceph_get_acl NULL
941 #define ceph_set_acl NULL
943 static inline int ceph_pre_init_acls(struct inode *dir, umode_t *mode,
944 struct ceph_acls_info *info)
948 static inline void ceph_init_inode_acls(struct inode *inode,
949 struct ceph_acls_info *info)
952 static inline void ceph_release_acls_info(struct ceph_acls_info *info)
955 static inline int ceph_acl_chmod(struct dentry *dentry, struct inode *inode)
960 static inline void ceph_forget_all_cached_acls(struct inode *inode)
967 extern const char *ceph_cap_string(int c);
968 extern void ceph_handle_caps(struct ceph_mds_session *session,
969 struct ceph_msg *msg);
970 extern struct ceph_cap *ceph_get_cap(struct ceph_mds_client *mdsc,
971 struct ceph_cap_reservation *ctx);
972 extern void ceph_add_cap(struct inode *inode,
973 struct ceph_mds_session *session, u64 cap_id,
974 int fmode, unsigned issued, unsigned wanted,
975 unsigned cap, unsigned seq, u64 realmino, int flags,
976 struct ceph_cap **new_cap);
977 extern void __ceph_remove_cap(struct ceph_cap *cap, bool queue_release);
978 extern void ceph_put_cap(struct ceph_mds_client *mdsc,
979 struct ceph_cap *cap);
980 extern int ceph_is_any_caps(struct inode *inode);
982 extern void ceph_queue_caps_release(struct inode *inode);
983 extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc);
984 extern int ceph_fsync(struct file *file, loff_t start, loff_t end,
986 extern void ceph_early_kick_flushing_caps(struct ceph_mds_client *mdsc,
987 struct ceph_mds_session *session);
988 extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc,
989 struct ceph_mds_session *session);
990 extern struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci,
992 extern int ceph_get_cap_mds(struct inode *inode);
993 extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps);
994 extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had);
995 extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
996 struct ceph_snap_context *snapc);
997 extern void ceph_flush_snaps(struct ceph_inode_info *ci,
998 struct ceph_mds_session **psession);
999 extern bool __ceph_should_report_size(struct ceph_inode_info *ci);
1000 extern void ceph_check_caps(struct ceph_inode_info *ci, int flags,
1001 struct ceph_mds_session *session);
1002 extern void ceph_check_delayed_caps(struct ceph_mds_client *mdsc);
1003 extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc);
1004 extern int ceph_drop_caps_for_unlink(struct inode *inode);
1005 extern int ceph_encode_inode_release(void **p, struct inode *inode,
1006 int mds, int drop, int unless, int force);
1007 extern int ceph_encode_dentry_release(void **p, struct dentry *dn,
1009 int mds, int drop, int unless);
1011 extern int ceph_get_caps(struct ceph_inode_info *ci, int need, int want,
1012 loff_t endoff, int *got, struct page **pinned_page);
1013 extern int ceph_try_get_caps(struct ceph_inode_info *ci,
1014 int need, int want, bool nonblock, int *got);
1016 /* for counting open files by mode */
1017 extern void __ceph_get_fmode(struct ceph_inode_info *ci, int mode);
1018 extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode);
1021 extern const struct address_space_operations ceph_aops;
1022 extern int ceph_mmap(struct file *file, struct vm_area_struct *vma);
1023 extern int ceph_uninline_data(struct file *filp, struct page *locked_page);
1024 extern int ceph_pool_perm_check(struct ceph_inode_info *ci, int need);
1025 extern void ceph_pool_perm_destroy(struct ceph_mds_client* mdsc);
1028 extern const struct file_operations ceph_file_fops;
1030 extern int ceph_renew_caps(struct inode *inode);
1031 extern int ceph_open(struct inode *inode, struct file *file);
1032 extern int ceph_atomic_open(struct inode *dir, struct dentry *dentry,
1033 struct file *file, unsigned flags, umode_t mode);
1034 extern int ceph_release(struct inode *inode, struct file *filp);
1035 extern void ceph_fill_inline_data(struct inode *inode, struct page *locked_page,
1036 char *data, size_t len);
1039 extern const struct file_operations ceph_dir_fops;
1040 extern const struct file_operations ceph_snapdir_fops;
1041 extern const struct inode_operations ceph_dir_iops;
1042 extern const struct inode_operations ceph_snapdir_iops;
1043 extern const struct dentry_operations ceph_dentry_ops;
1045 extern loff_t ceph_make_fpos(unsigned high, unsigned off, bool hash_order);
1046 extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry);
1047 extern int ceph_handle_snapdir(struct ceph_mds_request *req,
1048 struct dentry *dentry, int err);
1049 extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req,
1050 struct dentry *dentry, int err);
1052 extern void ceph_dentry_lru_add(struct dentry *dn);
1053 extern void ceph_dentry_lru_touch(struct dentry *dn);
1054 extern void ceph_dentry_lru_del(struct dentry *dn);
1055 extern void ceph_invalidate_dentry_lease(struct dentry *dentry);
1056 extern unsigned ceph_dentry_hash(struct inode *dir, struct dentry *dn);
1057 extern void ceph_readdir_cache_release(struct ceph_readdir_cache_control *ctl);
1060 extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
1063 extern const struct export_operations ceph_export_ops;
1066 extern __init void ceph_flock_init(void);
1067 extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl);
1068 extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl);
1069 extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num);
1070 extern int ceph_encode_locks_to_buffer(struct inode *inode,
1071 struct ceph_filelock *flocks,
1072 int num_fcntl_locks,
1073 int num_flock_locks);
1074 extern int ceph_locks_to_pagelist(struct ceph_filelock *flocks,
1075 struct ceph_pagelist *pagelist,
1076 int num_fcntl_locks, int num_flock_locks);
1079 extern int ceph_fs_debugfs_init(struct ceph_fs_client *client);
1080 extern void ceph_fs_debugfs_cleanup(struct ceph_fs_client *client);
1083 static inline bool __ceph_has_any_quota(struct ceph_inode_info *ci)
1085 return ci->i_max_files || ci->i_max_bytes;
1088 extern void ceph_adjust_quota_realms_count(struct inode *inode, bool inc);
1090 static inline void __ceph_update_quota(struct ceph_inode_info *ci,
1091 u64 max_bytes, u64 max_files)
1093 bool had_quota, has_quota;
1094 had_quota = __ceph_has_any_quota(ci);
1095 ci->i_max_bytes = max_bytes;
1096 ci->i_max_files = max_files;
1097 has_quota = __ceph_has_any_quota(ci);
1099 if (had_quota != has_quota)
1100 ceph_adjust_quota_realms_count(&ci->vfs_inode, has_quota);
1103 extern void ceph_handle_quota(struct ceph_mds_client *mdsc,
1104 struct ceph_mds_session *session,
1105 struct ceph_msg *msg);
1106 extern bool ceph_quota_is_max_files_exceeded(struct inode *inode);
1107 extern bool ceph_quota_is_same_realm(struct inode *old, struct inode *new);
1108 extern bool ceph_quota_is_max_bytes_exceeded(struct inode *inode,
1110 extern bool ceph_quota_is_max_bytes_approaching(struct inode *inode,
1112 extern bool ceph_quota_update_statfs(struct ceph_fs_client *fsc,
1113 struct kstatfs *buf);
1115 #endif /* _FS_CEPH_SUPER_H */