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 #define FSCACHE_USE_NEW_IO_API
25 #include <linux/fscache.h>
28 /* f_type in struct statfs */
29 #define CEPH_SUPER_MAGIC 0x00c36400
31 /* large granularity for statfs utilization stats to facilitate
32 * large volume sizes on 32-bit machines. */
33 #define CEPH_BLOCK_SHIFT 22 /* 4 MB */
34 #define CEPH_BLOCK (1 << CEPH_BLOCK_SHIFT)
36 #define CEPH_MOUNT_OPT_CLEANRECOVER (1<<1) /* auto reonnect (clean mode) after blocklisted */
37 #define CEPH_MOUNT_OPT_DIRSTAT (1<<4) /* `cat dirname` for stats */
38 #define CEPH_MOUNT_OPT_RBYTES (1<<5) /* dir st_bytes = rbytes */
39 #define CEPH_MOUNT_OPT_NOASYNCREADDIR (1<<7) /* no dcache readdir */
40 #define CEPH_MOUNT_OPT_INO32 (1<<8) /* 32 bit inos */
41 #define CEPH_MOUNT_OPT_DCACHE (1<<9) /* use dcache for readdir etc */
42 #define CEPH_MOUNT_OPT_FSCACHE (1<<10) /* use fscache */
43 #define CEPH_MOUNT_OPT_NOPOOLPERM (1<<11) /* no pool permission check */
44 #define CEPH_MOUNT_OPT_MOUNTWAIT (1<<12) /* mount waits if no mds is up */
45 #define CEPH_MOUNT_OPT_NOQUOTADF (1<<13) /* no root dir quota in statfs */
46 #define CEPH_MOUNT_OPT_NOCOPYFROM (1<<14) /* don't use RADOS 'copy-from' op */
47 #define CEPH_MOUNT_OPT_ASYNC_DIROPS (1<<15) /* allow async directory ops */
49 #define CEPH_MOUNT_OPT_DEFAULT \
50 (CEPH_MOUNT_OPT_DCACHE | \
51 CEPH_MOUNT_OPT_NOCOPYFROM | \
52 CEPH_MOUNT_OPT_ASYNC_DIROPS)
54 #define ceph_set_mount_opt(fsc, opt) \
55 (fsc)->mount_options->flags |= CEPH_MOUNT_OPT_##opt
56 #define ceph_clear_mount_opt(fsc, opt) \
57 (fsc)->mount_options->flags &= ~CEPH_MOUNT_OPT_##opt
58 #define ceph_test_mount_opt(fsc, opt) \
59 (!!((fsc)->mount_options->flags & CEPH_MOUNT_OPT_##opt))
61 /* max size of osd read request, limited by libceph */
62 #define CEPH_MAX_READ_SIZE CEPH_MSG_MAX_DATA_LEN
63 /* osd has a configurable 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_LEN
66 #define CEPH_RASIZE_DEFAULT (8192*1024) /* max readahead */
67 #define CEPH_MAX_READDIR_DEFAULT 1024
68 #define CEPH_MAX_READDIR_BYTES_DEFAULT (512*1024)
69 #define CEPH_SNAPDIRNAME_DEFAULT ".snap"
72 * Delay telling the MDS we no longer want caps, in case we reopen
73 * the file. Delay a minimum amount of time, even if we send a cap
74 * message for some other reason. Otherwise, take the oppotunity to
75 * update the mds to avoid sending another message later.
77 #define CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT 5 /* cap release delay */
78 #define CEPH_CAPS_WANTED_DELAY_MAX_DEFAULT 60 /* cap release delay */
80 struct ceph_mount_options {
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;
89 unsigned int max_readdir; /* max readdir result (entries) */
90 unsigned int max_readdir_bytes; /* max readdir result (bytes) */
93 * everything above this point can be memcmp'd; everything below
94 * is handled in compare_mount_options()
97 char *snapdir_name; /* default ".snap" */
98 char *mds_namespace; /* default NULL */
99 char *server_path; /* default NULL (means "/") */
100 char *fscache_uniq; /* default NULL */
103 struct ceph_fs_client {
104 struct super_block *sb;
106 struct list_head metric_wakeup;
108 struct ceph_mount_options *mount_options;
109 struct ceph_client *client;
115 bool have_copy_from2;
118 loff_t max_file_size;
120 struct ceph_mds_client *mdsc;
122 atomic_long_t writeback_count;
124 struct workqueue_struct *inode_wq;
125 struct workqueue_struct *cap_wq;
127 #ifdef CONFIG_DEBUG_FS
128 struct dentry *debugfs_dentry_lru, *debugfs_caps;
129 struct dentry *debugfs_congestion_kb;
130 struct dentry *debugfs_bdi;
131 struct dentry *debugfs_mdsc, *debugfs_mdsmap;
132 struct dentry *debugfs_metric;
133 struct dentry *debugfs_status;
134 struct dentry *debugfs_mds_sessions;
137 #ifdef CONFIG_CEPH_FSCACHE
138 struct fscache_cookie *fscache;
144 * File i/o capability. This tracks shared state with the metadata
145 * server that allows us to cache or writeback attributes or to read
146 * and write data. For any given inode, we should have one or more
147 * capabilities, one issued by each metadata server, and our
148 * cumulative access is the OR of all issued capabilities.
150 * Each cap is referenced by the inode's i_caps rbtree and by per-mds
151 * session capability lists.
154 struct ceph_inode_info *ci;
155 struct rb_node ci_node; /* per-ci cap tree */
156 struct ceph_mds_session *session;
157 struct list_head session_caps; /* per-session caplist */
158 u64 cap_id; /* unique cap id (mds provided) */
162 int issued; /* latest, from the mds */
163 int implemented; /* implemented superset of
164 issued (for revocation) */
165 int mds; /* mds index for this cap */
166 int mds_wanted; /* caps wanted from this mds */
168 /* caps to release */
174 u32 seq, issue_seq, mseq;
175 u32 cap_gen; /* active/stale cycle */
176 unsigned long last_used;
177 struct list_head caps_item;
180 #define CHECK_CAPS_AUTHONLY 1 /* only check auth cap */
181 #define CHECK_CAPS_FLUSH 2 /* flush any dirty caps */
182 #define CHECK_CAPS_NOINVAL 4 /* don't invalidate pagecache */
184 struct ceph_cap_flush {
187 bool wake; /* wake up flush waiters when finish ? */
188 bool is_capsnap; /* true means capsnap */
189 struct list_head g_list; // global
190 struct list_head i_list; // per inode
194 * Snapped cap state that is pending flush to mds. When a snapshot occurs,
195 * we first complete any in-process sync writes and writeback any dirty
196 * data before flushing the snapped state (tracked here) back to the MDS.
198 struct ceph_cap_snap {
200 struct list_head ci_item;
202 struct ceph_cap_flush cap_flush;
206 struct ceph_snap_context *context;
212 struct ceph_buffer *xattr_blob;
217 struct timespec64 mtime, atime, ctime, btime;
221 int writing; /* a sync write is still in progress */
222 int dirty_pages; /* dirty pages awaiting writeback */
227 static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap)
229 if (refcount_dec_and_test(&capsnap->nref)) {
230 if (capsnap->xattr_blob)
231 ceph_buffer_put(capsnap->xattr_blob);
237 * The frag tree describes how a directory is fragmented, potentially across
238 * multiple metadata servers. It is also used to indicate points where
239 * metadata authority is delegated, and whether/where metadata is replicated.
241 * A _leaf_ frag will be present in the i_fragtree IFF there is
242 * delegation info. That is, if mds >= 0 || ndist > 0.
244 #define CEPH_MAX_DIRFRAG_REP 4
246 struct ceph_inode_frag {
251 int split_by; /* i.e. 2^(split_by) children */
253 /* delegation and replication info */
254 int mds; /* -1 if same authority as parent */
255 int ndist; /* >0 if replicated */
256 int dist[CEPH_MAX_DIRFRAG_REP];
260 * We cache inode xattrs as an encoded blob until they are first used,
261 * at which point we parse them into an rbtree.
263 struct ceph_inode_xattr {
272 int should_free_name;
279 struct ceph_dentry_info {
280 struct dentry *dentry;
281 struct ceph_mds_session *lease_session;
282 struct list_head lease_list;
284 int lease_shared_gen;
287 unsigned long lease_renew_after, lease_renew_from;
292 #define CEPH_DENTRY_REFERENCED 1
293 #define CEPH_DENTRY_LEASE_LIST 2
294 #define CEPH_DENTRY_SHRINK_LIST 4
295 #define CEPH_DENTRY_PRIMARY_LINK 8
297 struct ceph_inode_xattrs_info {
299 * (still encoded) xattr blob. we avoid the overhead of parsing
300 * this until someone actually calls getxattr, etc.
302 * blob->vec.iov_len == 4 implies there are no xattrs; blob ==
303 * NULL means we don't know.
305 struct ceph_buffer *blob, *prealloc_blob;
307 struct rb_root index;
312 u64 version, index_version;
318 struct ceph_inode_info {
319 struct ceph_vino i_vino; /* ceph ino + snap */
321 spinlock_t i_ceph_lock;
324 u64 i_inline_version;
327 unsigned long i_ceph_flags;
328 atomic64_t i_release_count;
329 atomic64_t i_ordered_count;
330 atomic64_t i_complete_seq[2];
332 struct ceph_dir_layout i_dir_layout;
333 struct ceph_file_layout i_layout;
334 struct ceph_file_layout i_cached_layout; // for async creates
338 struct timespec64 i_rctime;
339 u64 i_rbytes, i_rfiles, i_rsubdirs, i_rsnaps;
340 u64 i_files, i_subdirs;
343 u64 i_max_bytes, i_max_files;
347 struct rb_root i_fragtree;
348 int i_fragtree_nsplits;
349 struct mutex i_fragtree_mutex;
351 struct ceph_inode_xattrs_info i_xattrs;
353 /* capabilities. protected _both_ by i_ceph_lock and cap->session's
355 struct rb_root i_caps; /* cap list */
356 struct ceph_cap *i_auth_cap; /* authoritative cap, if any */
357 unsigned i_dirty_caps, i_flushing_caps; /* mask of dirtied fields */
360 * Link to the auth cap's session's s_cap_dirty list. s_cap_dirty
361 * is protected by the mdsc->cap_dirty_lock, but each individual item
362 * is also protected by the inode's i_ceph_lock. Walking s_cap_dirty
363 * requires the mdsc->cap_dirty_lock. List presence for an item can
364 * be tested under the i_ceph_lock. Changing anything requires both.
366 struct list_head i_dirty_item;
369 * Link to session's s_cap_flushing list. Protected in a similar
370 * fashion to i_dirty_item, but also by the s_mutex for changes. The
371 * s_cap_flushing list can be walked while holding either the s_mutex
372 * or msdc->cap_dirty_lock. List presence can also be checked while
373 * holding the i_ceph_lock for this inode.
375 struct list_head i_flushing_item;
377 /* we need to track cap writeback on a per-cap-bit basis, to allow
378 * overlapping, pipelined cap flushes to the mds. we can probably
379 * reduce the tid to 8 bits if we're concerned about inode size. */
380 struct ceph_cap_flush *i_prealloc_cap_flush;
381 struct list_head i_cap_flush_list;
382 wait_queue_head_t i_cap_wq; /* threads waiting on a capability */
383 unsigned long i_hold_caps_max; /* jiffies */
384 struct list_head i_cap_delay_list; /* for delayed cap release to mds */
385 struct ceph_cap_reservation i_cap_migration_resv;
386 struct list_head i_cap_snaps; /* snapped state pending flush to mds */
387 struct ceph_snap_context *i_head_snapc; /* set if wr_buffer_head > 0 or
388 dirty|flushing caps */
389 unsigned i_snap_caps; /* cap bits for snapped files */
391 unsigned long i_last_rd;
392 unsigned long i_last_wr;
393 int i_nr_by_mode[CEPH_FILE_MODE_BITS]; /* open file counts */
395 struct mutex i_truncate_mutex;
396 u32 i_truncate_seq; /* last truncate to smaller size */
397 u64 i_truncate_size; /* and the size we last truncated down to */
398 int i_truncate_pending; /* still need to call vmtruncate */
400 u64 i_max_size; /* max file size authorized by mds */
401 u64 i_reported_size; /* (max_)size reported to or requested of mds */
402 u64 i_wanted_max_size; /* offset we'd like to write too */
403 u64 i_requested_max_size; /* max_size we've requested */
405 /* held references to caps */
407 int i_rd_ref, i_rdcache_ref, i_wr_ref, i_wb_ref, i_fx_ref;
408 int i_wrbuffer_ref, i_wrbuffer_ref_head;
409 atomic_t i_filelock_ref;
410 atomic_t i_shared_gen; /* increment each time we get FILE_SHARED */
411 u32 i_rdcache_gen; /* incremented each time we get FILE_CACHE. */
412 u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */
414 struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */
415 struct list_head i_unsafe_iops; /* uncommitted mds inode ops */
416 spinlock_t i_unsafe_lock;
419 struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */
420 struct ceph_snapid_map *i_snapid_map; /* snapid -> dev_t */
422 struct list_head i_snap_realm_item;
423 struct list_head i_snap_flush_item;
424 struct timespec64 i_btime;
425 struct timespec64 i_snap_btime;
427 struct work_struct i_work;
428 unsigned long i_work_mask;
430 #ifdef CONFIG_CEPH_FSCACHE
431 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 * The MDS reserves a set of inodes for its own usage. These should never
533 * be accessible by clients, and so the MDS has no reason to ever hand these
534 * out. The range is CEPH_MDS_INO_MDSDIR_OFFSET..CEPH_INO_SYSTEM_BASE.
536 * These come from src/mds/mdstypes.h in the ceph sources.
538 #define CEPH_MAX_MDS 0x100
539 #define CEPH_NUM_STRAY 10
540 #define CEPH_MDS_INO_MDSDIR_OFFSET (1 * CEPH_MAX_MDS)
541 #define CEPH_INO_SYSTEM_BASE ((6*CEPH_MAX_MDS) + (CEPH_MAX_MDS * CEPH_NUM_STRAY))
543 static inline bool ceph_vino_is_reserved(const struct ceph_vino vino)
545 if (vino.ino < CEPH_INO_SYSTEM_BASE &&
546 vino.ino >= CEPH_MDS_INO_MDSDIR_OFFSET) {
547 WARN_RATELIMIT(1, "Attempt to access reserved inode number 0x%llx", vino.ino);
553 static inline struct inode *ceph_find_inode(struct super_block *sb,
554 struct ceph_vino vino)
556 if (ceph_vino_is_reserved(vino))
560 * NB: The hashval will be run through the fs/inode.c hash function
561 * anyway, so there is no need to squash the inode number down to
562 * 32-bits first. Just use low-order bits on arches with 32-bit long.
564 return ilookup5(sb, (unsigned long)vino.ino, ceph_ino_compare, &vino);
571 #define CEPH_I_DIR_ORDERED (1 << 0) /* dentries in dir are ordered */
572 #define CEPH_I_FLUSH (1 << 2) /* do not delay flush of dirty metadata */
573 #define CEPH_I_POOL_PERM (1 << 3) /* pool rd/wr bits are valid */
574 #define CEPH_I_POOL_RD (1 << 4) /* can read from pool */
575 #define CEPH_I_POOL_WR (1 << 5) /* can write to pool */
576 #define CEPH_I_SEC_INITED (1 << 6) /* security initialized */
577 #define CEPH_I_KICK_FLUSH (1 << 7) /* kick flushing caps */
578 #define CEPH_I_FLUSH_SNAPS (1 << 8) /* need flush snapss */
579 #define CEPH_I_ERROR_WRITE (1 << 9) /* have seen write errors */
580 #define CEPH_I_ERROR_FILELOCK (1 << 10) /* have seen file lock errors */
581 #define CEPH_I_ODIRECT (1 << 11) /* inode in direct I/O mode */
582 #define CEPH_ASYNC_CREATE_BIT (12) /* async create in flight for this */
583 #define CEPH_I_ASYNC_CREATE (1 << CEPH_ASYNC_CREATE_BIT)
584 #define CEPH_I_SHUTDOWN (1 << 13) /* inode is no longer usable */
587 * Masks of ceph inode work.
589 #define CEPH_I_WORK_WRITEBACK 0
590 #define CEPH_I_WORK_INVALIDATE_PAGES 1
591 #define CEPH_I_WORK_VMTRUNCATE 2
592 #define CEPH_I_WORK_CHECK_CAPS 3
593 #define CEPH_I_WORK_FLUSH_SNAPS 4
596 * We set the ERROR_WRITE bit when we start seeing write errors on an inode
597 * and then clear it when they start succeeding. Note that we do a lockless
598 * check first, and only take the lock if it looks like it needs to be changed.
599 * The write submission code just takes this as a hint, so we're not too
600 * worried if a few slip through in either direction.
602 static inline void ceph_set_error_write(struct ceph_inode_info *ci)
604 if (!(READ_ONCE(ci->i_ceph_flags) & CEPH_I_ERROR_WRITE)) {
605 spin_lock(&ci->i_ceph_lock);
606 ci->i_ceph_flags |= CEPH_I_ERROR_WRITE;
607 spin_unlock(&ci->i_ceph_lock);
611 static inline void ceph_clear_error_write(struct ceph_inode_info *ci)
613 if (READ_ONCE(ci->i_ceph_flags) & CEPH_I_ERROR_WRITE) {
614 spin_lock(&ci->i_ceph_lock);
615 ci->i_ceph_flags &= ~CEPH_I_ERROR_WRITE;
616 spin_unlock(&ci->i_ceph_lock);
620 static inline void __ceph_dir_set_complete(struct ceph_inode_info *ci,
621 long long release_count,
622 long long ordered_count)
625 * Makes sure operations that setup readdir cache (update page
626 * cache and i_size) are strongly ordered w.r.t. the following
627 * atomic64_set() operations.
630 atomic64_set(&ci->i_complete_seq[0], release_count);
631 atomic64_set(&ci->i_complete_seq[1], ordered_count);
634 static inline void __ceph_dir_clear_complete(struct ceph_inode_info *ci)
636 atomic64_inc(&ci->i_release_count);
639 static inline void __ceph_dir_clear_ordered(struct ceph_inode_info *ci)
641 atomic64_inc(&ci->i_ordered_count);
644 static inline bool __ceph_dir_is_complete(struct ceph_inode_info *ci)
646 return atomic64_read(&ci->i_complete_seq[0]) ==
647 atomic64_read(&ci->i_release_count);
650 static inline bool __ceph_dir_is_complete_ordered(struct ceph_inode_info *ci)
652 return atomic64_read(&ci->i_complete_seq[0]) ==
653 atomic64_read(&ci->i_release_count) &&
654 atomic64_read(&ci->i_complete_seq[1]) ==
655 atomic64_read(&ci->i_ordered_count);
658 static inline void ceph_dir_clear_complete(struct inode *inode)
660 __ceph_dir_clear_complete(ceph_inode(inode));
663 static inline void ceph_dir_clear_ordered(struct inode *inode)
665 __ceph_dir_clear_ordered(ceph_inode(inode));
668 static inline bool ceph_dir_is_complete_ordered(struct inode *inode)
670 bool ret = __ceph_dir_is_complete_ordered(ceph_inode(inode));
675 /* find a specific frag @f */
676 extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci,
680 * choose fragment for value @v. copy frag content to pfrag, if leaf
683 extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v,
684 struct ceph_inode_frag *pfrag,
687 static inline struct ceph_dentry_info *ceph_dentry(const struct dentry *dentry)
689 return (struct ceph_dentry_info *)dentry->d_fsdata;
695 static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci)
697 return !RB_EMPTY_ROOT(&ci->i_caps);
700 extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented);
701 extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t);
702 extern int __ceph_caps_issued_mask_metric(struct ceph_inode_info *ci, int mask,
704 extern int __ceph_caps_issued_other(struct ceph_inode_info *ci,
705 struct ceph_cap *cap);
707 static inline int ceph_caps_issued(struct ceph_inode_info *ci)
710 spin_lock(&ci->i_ceph_lock);
711 issued = __ceph_caps_issued(ci, NULL);
712 spin_unlock(&ci->i_ceph_lock);
716 static inline int ceph_caps_issued_mask_metric(struct ceph_inode_info *ci,
720 spin_lock(&ci->i_ceph_lock);
721 r = __ceph_caps_issued_mask_metric(ci, mask, touch);
722 spin_unlock(&ci->i_ceph_lock);
726 static inline int __ceph_caps_dirty(struct ceph_inode_info *ci)
728 return ci->i_dirty_caps | ci->i_flushing_caps;
730 extern struct ceph_cap_flush *ceph_alloc_cap_flush(void);
731 extern void ceph_free_cap_flush(struct ceph_cap_flush *cf);
732 extern int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask,
733 struct ceph_cap_flush **pcf);
735 extern int __ceph_caps_revoking_other(struct ceph_inode_info *ci,
736 struct ceph_cap *ocap, int mask);
737 extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask);
738 extern int __ceph_caps_used(struct ceph_inode_info *ci);
740 static inline bool __ceph_is_file_opened(struct ceph_inode_info *ci)
742 return ci->i_nr_by_mode[0];
744 extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci);
745 extern int __ceph_caps_wanted(struct ceph_inode_info *ci);
747 /* what the mds thinks we want */
748 extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci, bool check);
750 extern void ceph_caps_init(struct ceph_mds_client *mdsc);
751 extern void ceph_caps_finalize(struct ceph_mds_client *mdsc);
752 extern void ceph_adjust_caps_max_min(struct ceph_mds_client *mdsc,
753 struct ceph_mount_options *fsopt);
754 extern int ceph_reserve_caps(struct ceph_mds_client *mdsc,
755 struct ceph_cap_reservation *ctx, int need);
756 extern void ceph_unreserve_caps(struct ceph_mds_client *mdsc,
757 struct ceph_cap_reservation *ctx);
758 extern void ceph_reservation_status(struct ceph_fs_client *client,
759 int *total, int *avail, int *used,
760 int *reserved, int *min);
765 * we keep buffered readdir results attached to file->private_data
767 #define CEPH_F_SYNC 1
768 #define CEPH_F_ATEND 2
770 struct ceph_file_info {
771 short fmode; /* initialized on open */
772 short flags; /* CEPH_F_* */
774 spinlock_t rw_contexts_lock;
775 struct list_head rw_contexts;
781 struct ceph_dir_file_info {
782 struct ceph_file_info file_info;
784 /* readdir: position within the dir */
786 struct ceph_mds_request *last_readdir;
788 /* readdir: position within a frag */
789 unsigned next_offset; /* offset of next chunk (last_name's + 1) */
790 char *last_name; /* last entry in previous chunk */
791 long long dir_release_count;
792 long long dir_ordered_count;
793 int readdir_cache_idx;
795 /* used for -o dirstat read() on directory thing */
800 struct ceph_rw_context {
801 struct list_head list;
802 struct task_struct *thread;
806 #define CEPH_DEFINE_RW_CONTEXT(_name, _caps) \
807 struct ceph_rw_context _name = { \
812 static inline void ceph_add_rw_context(struct ceph_file_info *cf,
813 struct ceph_rw_context *ctx)
815 spin_lock(&cf->rw_contexts_lock);
816 list_add(&ctx->list, &cf->rw_contexts);
817 spin_unlock(&cf->rw_contexts_lock);
820 static inline void ceph_del_rw_context(struct ceph_file_info *cf,
821 struct ceph_rw_context *ctx)
823 spin_lock(&cf->rw_contexts_lock);
824 list_del(&ctx->list);
825 spin_unlock(&cf->rw_contexts_lock);
828 static inline struct ceph_rw_context*
829 ceph_find_rw_context(struct ceph_file_info *cf)
831 struct ceph_rw_context *ctx, *found = NULL;
832 spin_lock(&cf->rw_contexts_lock);
833 list_for_each_entry(ctx, &cf->rw_contexts, list) {
834 if (ctx->thread == current) {
839 spin_unlock(&cf->rw_contexts_lock);
843 struct ceph_readdir_cache_control {
845 struct dentry **dentries;
850 * A "snap realm" describes a subset of the file hierarchy sharing
851 * the same set of snapshots that apply to it. The realms themselves
852 * are organized into a hierarchy, such that children inherit (some of)
853 * the snapshots of their parents.
855 * All inodes within the realm that have capabilities are linked into a
858 struct ceph_snap_realm {
866 u64 parent_since; /* snapid when our current parent became so */
868 u64 *prior_parent_snaps; /* snaps inherited from any parents we */
869 u32 num_prior_parent_snaps; /* had prior to parent_since */
870 u64 *snaps; /* snaps specific to this realm */
873 struct ceph_snap_realm *parent;
874 struct list_head children; /* list of child realms */
875 struct list_head child_item;
877 struct list_head empty_item; /* if i have ref==0 */
879 struct list_head dirty_item; /* if realm needs new context */
881 /* the current set of snaps for this realm */
882 struct ceph_snap_context *cached_context;
884 struct list_head inodes_with_caps;
885 spinlock_t inodes_with_caps_lock;
888 static inline int default_congestion_kb(void)
895 * congestion size, scale with available memory.
907 * This allows larger machines to have larger/more transfers.
908 * Limit the default to 256M
910 congestion_kb = (16*int_sqrt(totalram_pages())) << (PAGE_SHIFT-10);
911 if (congestion_kb > 256*1024)
912 congestion_kb = 256*1024;
914 return congestion_kb;
919 extern int ceph_force_reconnect(struct super_block *sb);
921 struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc,
923 extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc,
924 struct ceph_snap_realm *realm);
925 extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc,
926 struct ceph_snap_realm *realm);
927 extern int ceph_update_snap_trace(struct ceph_mds_client *m,
928 void *p, void *e, bool deletion,
929 struct ceph_snap_realm **realm_ret);
930 void ceph_change_snap_realm(struct inode *inode, struct ceph_snap_realm *realm);
931 extern void ceph_handle_snap(struct ceph_mds_client *mdsc,
932 struct ceph_mds_session *session,
933 struct ceph_msg *msg);
934 extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci,
935 struct ceph_cap_snap *capsnap);
936 extern void ceph_cleanup_empty_realms(struct ceph_mds_client *mdsc);
938 extern struct ceph_snapid_map *ceph_get_snapid_map(struct ceph_mds_client *mdsc,
940 extern void ceph_put_snapid_map(struct ceph_mds_client* mdsc,
941 struct ceph_snapid_map *sm);
942 extern void ceph_trim_snapid_map(struct ceph_mds_client *mdsc);
943 extern void ceph_cleanup_snapid_map(struct ceph_mds_client *mdsc);
947 * a cap_snap is "pending" if it is still awaiting an in-progress
948 * sync write (that may/may not still update size, mtime, etc.).
950 static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci)
952 return !list_empty(&ci->i_cap_snaps) &&
953 list_last_entry(&ci->i_cap_snaps, struct ceph_cap_snap,
958 struct ceph_mds_reply_info_in;
959 struct ceph_mds_reply_dirfrag;
961 extern const struct inode_operations ceph_file_iops;
963 extern struct inode *ceph_alloc_inode(struct super_block *sb);
964 extern void ceph_evict_inode(struct inode *inode);
965 extern void ceph_free_inode(struct inode *inode);
967 extern struct inode *ceph_get_inode(struct super_block *sb,
968 struct ceph_vino vino);
969 extern struct inode *ceph_get_snapdir(struct inode *parent);
970 extern int ceph_fill_file_size(struct inode *inode, int issued,
971 u32 truncate_seq, u64 truncate_size, u64 size);
972 extern void ceph_fill_file_time(struct inode *inode, int issued,
973 u64 time_warp_seq, struct timespec64 *ctime,
974 struct timespec64 *mtime,
975 struct timespec64 *atime);
976 extern int ceph_fill_inode(struct inode *inode, struct page *locked_page,
977 struct ceph_mds_reply_info_in *iinfo,
978 struct ceph_mds_reply_dirfrag *dirinfo,
979 struct ceph_mds_session *session, int cap_fmode,
980 struct ceph_cap_reservation *caps_reservation);
981 extern int ceph_fill_trace(struct super_block *sb,
982 struct ceph_mds_request *req);
983 extern int ceph_readdir_prepopulate(struct ceph_mds_request *req,
984 struct ceph_mds_session *session);
986 extern int ceph_inode_holds_cap(struct inode *inode, int mask);
988 extern bool ceph_inode_set_size(struct inode *inode, loff_t size);
989 extern void __ceph_do_pending_vmtruncate(struct inode *inode);
991 void ceph_queue_inode_work(struct inode *inode, int work_bit);
993 static inline void ceph_queue_vmtruncate(struct inode *inode)
995 ceph_queue_inode_work(inode, CEPH_I_WORK_VMTRUNCATE);
998 static inline void ceph_queue_invalidate(struct inode *inode)
1000 ceph_queue_inode_work(inode, CEPH_I_WORK_INVALIDATE_PAGES);
1003 static inline void ceph_queue_writeback(struct inode *inode)
1005 ceph_queue_inode_work(inode, CEPH_I_WORK_WRITEBACK);
1008 static inline void ceph_queue_check_caps(struct inode *inode)
1010 ceph_queue_inode_work(inode, CEPH_I_WORK_CHECK_CAPS);
1013 static inline void ceph_queue_flush_snaps(struct inode *inode)
1015 ceph_queue_inode_work(inode, CEPH_I_WORK_FLUSH_SNAPS);
1018 extern int __ceph_do_getattr(struct inode *inode, struct page *locked_page,
1019 int mask, bool force);
1020 static inline int ceph_do_getattr(struct inode *inode, int mask, bool force)
1022 return __ceph_do_getattr(inode, NULL, mask, force);
1024 extern int ceph_permission(struct user_namespace *mnt_userns,
1025 struct inode *inode, int mask);
1026 extern int __ceph_setattr(struct inode *inode, struct iattr *attr);
1027 extern int ceph_setattr(struct user_namespace *mnt_userns,
1028 struct dentry *dentry, struct iattr *attr);
1029 extern int ceph_getattr(struct user_namespace *mnt_userns,
1030 const struct path *path, struct kstat *stat,
1031 u32 request_mask, unsigned int flags);
1032 void ceph_inode_shutdown(struct inode *inode);
1034 static inline bool ceph_inode_is_shutdown(struct inode *inode)
1036 unsigned long flags = READ_ONCE(ceph_inode(inode)->i_ceph_flags);
1037 struct ceph_fs_client *fsc = ceph_inode_to_client(inode);
1038 int state = READ_ONCE(fsc->mount_state);
1040 return (flags & CEPH_I_SHUTDOWN) || state >= CEPH_MOUNT_SHUTDOWN;
1044 int __ceph_setxattr(struct inode *, const char *, const void *, size_t, int);
1045 ssize_t __ceph_getxattr(struct inode *, const char *, void *, size_t);
1046 extern ssize_t ceph_listxattr(struct dentry *, char *, size_t);
1047 extern struct ceph_buffer *__ceph_build_xattrs_blob(struct ceph_inode_info *ci);
1048 extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci);
1049 extern const struct xattr_handler *ceph_xattr_handlers[];
1051 struct ceph_acl_sec_ctx {
1052 #ifdef CONFIG_CEPH_FS_POSIX_ACL
1056 #ifdef CONFIG_CEPH_FS_SECURITY_LABEL
1060 struct ceph_pagelist *pagelist;
1063 #ifdef CONFIG_SECURITY
1064 extern bool ceph_security_xattr_deadlock(struct inode *in);
1065 extern bool ceph_security_xattr_wanted(struct inode *in);
1067 static inline bool ceph_security_xattr_deadlock(struct inode *in)
1071 static inline bool ceph_security_xattr_wanted(struct inode *in)
1077 #ifdef CONFIG_CEPH_FS_SECURITY_LABEL
1078 extern int ceph_security_init_secctx(struct dentry *dentry, umode_t mode,
1079 struct ceph_acl_sec_ctx *ctx);
1080 static inline void ceph_security_invalidate_secctx(struct inode *inode)
1082 security_inode_invalidate_secctx(inode);
1085 static inline int ceph_security_init_secctx(struct dentry *dentry, umode_t mode,
1086 struct ceph_acl_sec_ctx *ctx)
1090 static inline void ceph_security_invalidate_secctx(struct inode *inode)
1095 void ceph_release_acl_sec_ctx(struct ceph_acl_sec_ctx *as_ctx);
1098 #ifdef CONFIG_CEPH_FS_POSIX_ACL
1100 struct posix_acl *ceph_get_acl(struct inode *, int, bool);
1101 int ceph_set_acl(struct user_namespace *mnt_userns,
1102 struct inode *inode, struct posix_acl *acl, int type);
1103 int ceph_pre_init_acls(struct inode *dir, umode_t *mode,
1104 struct ceph_acl_sec_ctx *as_ctx);
1105 void ceph_init_inode_acls(struct inode *inode,
1106 struct ceph_acl_sec_ctx *as_ctx);
1108 static inline void ceph_forget_all_cached_acls(struct inode *inode)
1110 forget_all_cached_acls(inode);
1115 #define ceph_get_acl NULL
1116 #define ceph_set_acl NULL
1118 static inline int ceph_pre_init_acls(struct inode *dir, umode_t *mode,
1119 struct ceph_acl_sec_ctx *as_ctx)
1123 static inline void ceph_init_inode_acls(struct inode *inode,
1124 struct ceph_acl_sec_ctx *as_ctx)
1127 static inline int ceph_acl_chmod(struct dentry *dentry, struct inode *inode)
1132 static inline void ceph_forget_all_cached_acls(struct inode *inode)
1139 extern const char *ceph_cap_string(int c);
1140 extern void ceph_handle_caps(struct ceph_mds_session *session,
1141 struct ceph_msg *msg);
1142 extern struct ceph_cap *ceph_get_cap(struct ceph_mds_client *mdsc,
1143 struct ceph_cap_reservation *ctx);
1144 extern void ceph_add_cap(struct inode *inode,
1145 struct ceph_mds_session *session, u64 cap_id,
1146 unsigned issued, unsigned wanted,
1147 unsigned cap, unsigned seq, u64 realmino, int flags,
1148 struct ceph_cap **new_cap);
1149 extern void __ceph_remove_cap(struct ceph_cap *cap, bool queue_release);
1150 extern void ceph_remove_cap(struct ceph_cap *cap, bool queue_release);
1151 extern void __ceph_remove_caps(struct ceph_inode_info *ci);
1152 extern void ceph_put_cap(struct ceph_mds_client *mdsc,
1153 struct ceph_cap *cap);
1154 extern int ceph_is_any_caps(struct inode *inode);
1156 extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc);
1157 extern int ceph_fsync(struct file *file, loff_t start, loff_t end,
1159 extern void ceph_early_kick_flushing_caps(struct ceph_mds_client *mdsc,
1160 struct ceph_mds_session *session);
1161 extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc,
1162 struct ceph_mds_session *session);
1163 void ceph_kick_flushing_inode_caps(struct ceph_mds_session *session,
1164 struct ceph_inode_info *ci);
1165 extern struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci,
1167 extern void ceph_take_cap_refs(struct ceph_inode_info *ci, int caps,
1168 bool snap_rwsem_locked);
1169 extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps);
1170 extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had);
1171 extern void ceph_put_cap_refs_async(struct ceph_inode_info *ci, int had);
1172 extern void ceph_put_cap_refs_no_check_caps(struct ceph_inode_info *ci,
1174 extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
1175 struct ceph_snap_context *snapc);
1176 extern void __ceph_remove_capsnap(struct inode *inode,
1177 struct ceph_cap_snap *capsnap,
1178 bool *wake_ci, bool *wake_mdsc);
1179 extern void ceph_remove_capsnap(struct inode *inode,
1180 struct ceph_cap_snap *capsnap,
1181 bool *wake_ci, bool *wake_mdsc);
1182 extern void ceph_flush_snaps(struct ceph_inode_info *ci,
1183 struct ceph_mds_session **psession);
1184 extern bool __ceph_should_report_size(struct ceph_inode_info *ci);
1185 extern void ceph_check_caps(struct ceph_inode_info *ci, int flags,
1186 struct ceph_mds_session *session);
1187 extern unsigned long ceph_check_delayed_caps(struct ceph_mds_client *mdsc);
1188 extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc);
1189 extern int ceph_drop_caps_for_unlink(struct inode *inode);
1190 extern int ceph_encode_inode_release(void **p, struct inode *inode,
1191 int mds, int drop, int unless, int force);
1192 extern int ceph_encode_dentry_release(void **p, struct dentry *dn,
1194 int mds, int drop, int unless);
1196 extern int ceph_get_caps(struct file *filp, int need, int want,
1197 loff_t endoff, int *got);
1198 extern int ceph_try_get_caps(struct inode *inode,
1199 int need, int want, bool nonblock, int *got);
1201 /* for counting open files by mode */
1202 extern void ceph_get_fmode(struct ceph_inode_info *ci, int mode, int count);
1203 extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode, int count);
1204 extern void __ceph_touch_fmode(struct ceph_inode_info *ci,
1205 struct ceph_mds_client *mdsc, int fmode);
1208 extern const struct address_space_operations ceph_aops;
1209 extern int ceph_mmap(struct file *file, struct vm_area_struct *vma);
1210 extern int ceph_uninline_data(struct file *filp, struct page *locked_page);
1211 extern int ceph_pool_perm_check(struct inode *inode, int need);
1212 extern void ceph_pool_perm_destroy(struct ceph_mds_client* mdsc);
1213 int ceph_purge_inode_cap(struct inode *inode, struct ceph_cap *cap, bool *invalidate);
1216 extern const struct file_operations ceph_file_fops;
1218 extern int ceph_renew_caps(struct inode *inode, int fmode);
1219 extern int ceph_open(struct inode *inode, struct file *file);
1220 extern int ceph_atomic_open(struct inode *dir, struct dentry *dentry,
1221 struct file *file, unsigned flags, umode_t mode);
1222 extern int ceph_release(struct inode *inode, struct file *filp);
1223 extern void ceph_fill_inline_data(struct inode *inode, struct page *locked_page,
1224 char *data, size_t len);
1227 extern const struct file_operations ceph_dir_fops;
1228 extern const struct file_operations ceph_snapdir_fops;
1229 extern const struct inode_operations ceph_dir_iops;
1230 extern const struct inode_operations ceph_snapdir_iops;
1231 extern const struct dentry_operations ceph_dentry_ops;
1233 extern loff_t ceph_make_fpos(unsigned high, unsigned off, bool hash_order);
1234 extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry);
1235 extern struct dentry *ceph_handle_snapdir(struct ceph_mds_request *req,
1236 struct dentry *dentry);
1237 extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req,
1238 struct dentry *dentry, int err);
1240 extern void __ceph_dentry_lease_touch(struct ceph_dentry_info *di);
1241 extern void __ceph_dentry_dir_lease_touch(struct ceph_dentry_info *di);
1242 extern void ceph_invalidate_dentry_lease(struct dentry *dentry);
1243 extern int ceph_trim_dentries(struct ceph_mds_client *mdsc);
1244 extern unsigned ceph_dentry_hash(struct inode *dir, struct dentry *dn);
1245 extern void ceph_readdir_cache_release(struct ceph_readdir_cache_control *ctl);
1248 extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
1251 extern const struct export_operations ceph_export_ops;
1252 struct inode *ceph_lookup_inode(struct super_block *sb, u64 ino);
1255 extern __init void ceph_flock_init(void);
1256 extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl);
1257 extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl);
1258 extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num);
1259 extern int ceph_encode_locks_to_buffer(struct inode *inode,
1260 struct ceph_filelock *flocks,
1261 int num_fcntl_locks,
1262 int num_flock_locks);
1263 extern int ceph_locks_to_pagelist(struct ceph_filelock *flocks,
1264 struct ceph_pagelist *pagelist,
1265 int num_fcntl_locks, int num_flock_locks);
1268 extern void ceph_fs_debugfs_init(struct ceph_fs_client *client);
1269 extern void ceph_fs_debugfs_cleanup(struct ceph_fs_client *client);
1272 static inline bool __ceph_has_any_quota(struct ceph_inode_info *ci)
1274 return ci->i_max_files || ci->i_max_bytes;
1277 extern void ceph_adjust_quota_realms_count(struct inode *inode, bool inc);
1279 static inline void __ceph_update_quota(struct ceph_inode_info *ci,
1280 u64 max_bytes, u64 max_files)
1282 bool had_quota, has_quota;
1283 had_quota = __ceph_has_any_quota(ci);
1284 ci->i_max_bytes = max_bytes;
1285 ci->i_max_files = max_files;
1286 has_quota = __ceph_has_any_quota(ci);
1288 if (had_quota != has_quota)
1289 ceph_adjust_quota_realms_count(&ci->vfs_inode, has_quota);
1292 extern void ceph_handle_quota(struct ceph_mds_client *mdsc,
1293 struct ceph_mds_session *session,
1294 struct ceph_msg *msg);
1295 extern bool ceph_quota_is_max_files_exceeded(struct inode *inode);
1296 extern bool ceph_quota_is_same_realm(struct inode *old, struct inode *new);
1297 extern bool ceph_quota_is_max_bytes_exceeded(struct inode *inode,
1299 extern bool ceph_quota_is_max_bytes_approaching(struct inode *inode,
1301 extern bool ceph_quota_update_statfs(struct ceph_fs_client *fsc,
1302 struct kstatfs *buf);
1303 extern void ceph_cleanup_quotarealms_inodes(struct ceph_mds_client *mdsc);
1305 #endif /* _FS_CEPH_SUPER_H */