Loading...
1#ifndef _FS_CEPH_SUPER_H
2#define _FS_CEPH_SUPER_H
3
4#include <linux/ceph/ceph_debug.h>
5
6#include <asm/unaligned.h>
7#include <linux/backing-dev.h>
8#include <linux/completion.h>
9#include <linux/exportfs.h>
10#include <linux/fs.h>
11#include <linux/mempool.h>
12#include <linux/pagemap.h>
13#include <linux/wait.h>
14#include <linux/writeback.h>
15#include <linux/slab.h>
16#include <linux/posix_acl.h>
17
18#include <linux/ceph/libceph.h>
19
20#ifdef CONFIG_CEPH_FSCACHE
21#include <linux/fscache.h>
22#endif
23
24/* f_type in struct statfs */
25#define CEPH_SUPER_MAGIC 0x00c36400
26
27/* large granularity for statfs utilization stats to facilitate
28 * large volume sizes on 32-bit machines. */
29#define CEPH_BLOCK_SHIFT 22 /* 4 MB */
30#define CEPH_BLOCK (1 << CEPH_BLOCK_SHIFT)
31
32#define CEPH_MOUNT_OPT_DIRSTAT (1<<4) /* `cat dirname` for stats */
33#define CEPH_MOUNT_OPT_RBYTES (1<<5) /* dir st_bytes = rbytes */
34#define CEPH_MOUNT_OPT_NOASYNCREADDIR (1<<7) /* no dcache readdir */
35#define CEPH_MOUNT_OPT_INO32 (1<<8) /* 32 bit inos */
36#define CEPH_MOUNT_OPT_DCACHE (1<<9) /* use dcache for readdir etc */
37#define CEPH_MOUNT_OPT_FSCACHE (1<<10) /* use fscache */
38#define CEPH_MOUNT_OPT_NOPOOLPERM (1<<11) /* no pool permission check */
39
40#define CEPH_MOUNT_OPT_DEFAULT CEPH_MOUNT_OPT_DCACHE
41
42#define ceph_set_mount_opt(fsc, opt) \
43 (fsc)->mount_options->flags |= CEPH_MOUNT_OPT_##opt;
44#define ceph_test_mount_opt(fsc, opt) \
45 (!!((fsc)->mount_options->flags & CEPH_MOUNT_OPT_##opt))
46
47#define CEPH_RSIZE_DEFAULT 0 /* max read size */
48#define CEPH_RASIZE_DEFAULT (8192*1024) /* readahead */
49#define CEPH_MAX_READDIR_DEFAULT 1024
50#define CEPH_MAX_READDIR_BYTES_DEFAULT (512*1024)
51#define CEPH_SNAPDIRNAME_DEFAULT ".snap"
52
53struct ceph_mount_options {
54 int flags;
55 int sb_flags;
56
57 int wsize; /* max write size */
58 int rsize; /* max read size */
59 int rasize; /* max readahead */
60 int congestion_kb; /* max writeback in flight */
61 int caps_wanted_delay_min, caps_wanted_delay_max;
62 int cap_release_safety;
63 int max_readdir; /* max readdir result (entires) */
64 int max_readdir_bytes; /* max readdir result (bytes) */
65
66 /*
67 * everything above this point can be memcmp'd; everything below
68 * is handled in compare_mount_options()
69 */
70
71 char *snapdir_name; /* default ".snap" */
72};
73
74struct ceph_fs_client {
75 struct super_block *sb;
76
77 struct ceph_mount_options *mount_options;
78 struct ceph_client *client;
79
80 unsigned long mount_state;
81 int min_caps; /* min caps i added */
82
83 struct ceph_mds_client *mdsc;
84
85 /* writeback */
86 mempool_t *wb_pagevec_pool;
87 struct workqueue_struct *wb_wq;
88 struct workqueue_struct *pg_inv_wq;
89 struct workqueue_struct *trunc_wq;
90 atomic_long_t writeback_count;
91
92 struct backing_dev_info backing_dev_info;
93
94#ifdef CONFIG_DEBUG_FS
95 struct dentry *debugfs_dentry_lru, *debugfs_caps;
96 struct dentry *debugfs_congestion_kb;
97 struct dentry *debugfs_bdi;
98 struct dentry *debugfs_mdsc, *debugfs_mdsmap;
99 struct dentry *debugfs_mds_sessions;
100#endif
101
102#ifdef CONFIG_CEPH_FSCACHE
103 struct fscache_cookie *fscache;
104 struct workqueue_struct *revalidate_wq;
105#endif
106};
107
108
109/*
110 * File i/o capability. This tracks shared state with the metadata
111 * server that allows us to cache or writeback attributes or to read
112 * and write data. For any given inode, we should have one or more
113 * capabilities, one issued by each metadata server, and our
114 * cumulative access is the OR of all issued capabilities.
115 *
116 * Each cap is referenced by the inode's i_caps rbtree and by per-mds
117 * session capability lists.
118 */
119struct ceph_cap {
120 struct ceph_inode_info *ci;
121 struct rb_node ci_node; /* per-ci cap tree */
122 struct ceph_mds_session *session;
123 struct list_head session_caps; /* per-session caplist */
124 u64 cap_id; /* unique cap id (mds provided) */
125 union {
126 /* in-use caps */
127 struct {
128 int issued; /* latest, from the mds */
129 int implemented; /* implemented superset of
130 issued (for revocation) */
131 int mds, mds_wanted;
132 };
133 /* caps to release */
134 struct {
135 u64 cap_ino;
136 int queue_release;
137 };
138 };
139 u32 seq, issue_seq, mseq;
140 u32 cap_gen; /* active/stale cycle */
141 unsigned long last_used;
142 struct list_head caps_item;
143};
144
145#define CHECK_CAPS_NODELAY 1 /* do not delay any further */
146#define CHECK_CAPS_AUTHONLY 2 /* only check auth cap */
147#define CHECK_CAPS_FLUSH 4 /* flush any dirty caps */
148
149/*
150 * Snapped cap state that is pending flush to mds. When a snapshot occurs,
151 * we first complete any in-process sync writes and writeback any dirty
152 * data before flushing the snapped state (tracked here) back to the MDS.
153 */
154struct ceph_cap_snap {
155 atomic_t nref;
156 struct ceph_inode_info *ci;
157 struct list_head ci_item, flushing_item;
158
159 u64 follows, flush_tid;
160 int issued, dirty;
161 struct ceph_snap_context *context;
162
163 umode_t mode;
164 kuid_t uid;
165 kgid_t gid;
166
167 struct ceph_buffer *xattr_blob;
168 u64 xattr_version;
169
170 u64 size;
171 struct timespec mtime, atime, ctime;
172 u64 time_warp_seq;
173 int writing; /* a sync write is still in progress */
174 int dirty_pages; /* dirty pages awaiting writeback */
175 bool inline_data;
176 bool need_flush;
177};
178
179static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap)
180{
181 if (atomic_dec_and_test(&capsnap->nref)) {
182 if (capsnap->xattr_blob)
183 ceph_buffer_put(capsnap->xattr_blob);
184 kfree(capsnap);
185 }
186}
187
188struct ceph_cap_flush {
189 u64 tid;
190 int caps;
191 struct rb_node g_node; // global
192 union {
193 struct rb_node i_node; // inode
194 struct list_head list;
195 };
196};
197
198/*
199 * The frag tree describes how a directory is fragmented, potentially across
200 * multiple metadata servers. It is also used to indicate points where
201 * metadata authority is delegated, and whether/where metadata is replicated.
202 *
203 * A _leaf_ frag will be present in the i_fragtree IFF there is
204 * delegation info. That is, if mds >= 0 || ndist > 0.
205 */
206#define CEPH_MAX_DIRFRAG_REP 4
207
208struct ceph_inode_frag {
209 struct rb_node node;
210
211 /* fragtree state */
212 u32 frag;
213 int split_by; /* i.e. 2^(split_by) children */
214
215 /* delegation and replication info */
216 int mds; /* -1 if same authority as parent */
217 int ndist; /* >0 if replicated */
218 int dist[CEPH_MAX_DIRFRAG_REP];
219};
220
221/*
222 * We cache inode xattrs as an encoded blob until they are first used,
223 * at which point we parse them into an rbtree.
224 */
225struct ceph_inode_xattr {
226 struct rb_node node;
227
228 const char *name;
229 int name_len;
230 const char *val;
231 int val_len;
232 int dirty;
233
234 int should_free_name;
235 int should_free_val;
236};
237
238/*
239 * Ceph dentry state
240 */
241struct ceph_dentry_info {
242 struct ceph_mds_session *lease_session;
243 u32 lease_gen, lease_shared_gen;
244 u32 lease_seq;
245 unsigned long lease_renew_after, lease_renew_from;
246 struct list_head lru;
247 struct dentry *dentry;
248 u64 time;
249 u64 offset;
250};
251
252struct ceph_inode_xattrs_info {
253 /*
254 * (still encoded) xattr blob. we avoid the overhead of parsing
255 * this until someone actually calls getxattr, etc.
256 *
257 * blob->vec.iov_len == 4 implies there are no xattrs; blob ==
258 * NULL means we don't know.
259 */
260 struct ceph_buffer *blob, *prealloc_blob;
261
262 struct rb_root index;
263 bool dirty;
264 int count;
265 int names_size;
266 int vals_size;
267 u64 version, index_version;
268};
269
270/*
271 * Ceph inode.
272 */
273struct ceph_inode_info {
274 struct ceph_vino i_vino; /* ceph ino + snap */
275
276 spinlock_t i_ceph_lock;
277
278 u64 i_version;
279 u64 i_inline_version;
280 u32 i_time_warp_seq;
281
282 unsigned i_ceph_flags;
283 atomic64_t i_release_count;
284 atomic64_t i_ordered_count;
285 atomic64_t i_complete_seq[2];
286
287 struct ceph_dir_layout i_dir_layout;
288 struct ceph_file_layout i_layout;
289 size_t i_pool_ns_len;
290 char *i_symlink;
291
292 /* for dirs */
293 struct timespec i_rctime;
294 u64 i_rbytes, i_rfiles, i_rsubdirs;
295 u64 i_files, i_subdirs;
296
297 struct rb_root i_fragtree;
298 struct mutex i_fragtree_mutex;
299
300 struct ceph_inode_xattrs_info i_xattrs;
301
302 /* capabilities. protected _both_ by i_ceph_lock and cap->session's
303 * s_mutex. */
304 struct rb_root i_caps; /* cap list */
305 struct ceph_cap *i_auth_cap; /* authoritative cap, if any */
306 unsigned i_dirty_caps, i_flushing_caps; /* mask of dirtied fields */
307 struct list_head i_dirty_item, i_flushing_item;
308 /* we need to track cap writeback on a per-cap-bit basis, to allow
309 * overlapping, pipelined cap flushes to the mds. we can probably
310 * reduce the tid to 8 bits if we're concerned about inode size. */
311 struct ceph_cap_flush *i_prealloc_cap_flush;
312 struct rb_root i_cap_flush_tree;
313 wait_queue_head_t i_cap_wq; /* threads waiting on a capability */
314 unsigned long i_hold_caps_min; /* jiffies */
315 unsigned long i_hold_caps_max; /* jiffies */
316 struct list_head i_cap_delay_list; /* for delayed cap release to mds */
317 struct ceph_cap_reservation i_cap_migration_resv;
318 struct list_head i_cap_snaps; /* snapped state pending flush to mds */
319 struct ceph_snap_context *i_head_snapc; /* set if wr_buffer_head > 0 or
320 dirty|flushing caps */
321 unsigned i_snap_caps; /* cap bits for snapped files */
322
323 int i_nr_by_mode[CEPH_FILE_MODE_NUM]; /* open file counts */
324
325 struct mutex i_truncate_mutex;
326 u32 i_truncate_seq; /* last truncate to smaller size */
327 u64 i_truncate_size; /* and the size we last truncated down to */
328 int i_truncate_pending; /* still need to call vmtruncate */
329
330 u64 i_max_size; /* max file size authorized by mds */
331 u64 i_reported_size; /* (max_)size reported to or requested of mds */
332 u64 i_wanted_max_size; /* offset we'd like to write too */
333 u64 i_requested_max_size; /* max_size we've requested */
334
335 /* held references to caps */
336 int i_pin_ref;
337 int i_rd_ref, i_rdcache_ref, i_wr_ref, i_wb_ref;
338 int i_wrbuffer_ref, i_wrbuffer_ref_head;
339 u32 i_shared_gen; /* increment each time we get FILE_SHARED */
340 u32 i_rdcache_gen; /* incremented each time we get FILE_CACHE. */
341 u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */
342
343 struct list_head i_unsafe_writes; /* uncommitted sync writes */
344 struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */
345 struct list_head i_unsafe_iops; /* uncommitted mds inode ops */
346 spinlock_t i_unsafe_lock;
347
348 struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */
349 int i_snap_realm_counter; /* snap realm (if caps) */
350 struct list_head i_snap_realm_item;
351 struct list_head i_snap_flush_item;
352
353 struct work_struct i_wb_work; /* writeback work */
354 struct work_struct i_pg_inv_work; /* page invalidation work */
355
356 struct work_struct i_vmtruncate_work;
357
358#ifdef CONFIG_CEPH_FSCACHE
359 struct fscache_cookie *fscache;
360 u32 i_fscache_gen; /* sequence, for delayed fscache validate */
361 struct work_struct i_revalidate_work;
362#endif
363 struct inode vfs_inode; /* at end */
364};
365
366static inline struct ceph_inode_info *ceph_inode(struct inode *inode)
367{
368 return container_of(inode, struct ceph_inode_info, vfs_inode);
369}
370
371static inline struct ceph_fs_client *ceph_inode_to_client(struct inode *inode)
372{
373 return (struct ceph_fs_client *)inode->i_sb->s_fs_info;
374}
375
376static inline struct ceph_fs_client *ceph_sb_to_client(struct super_block *sb)
377{
378 return (struct ceph_fs_client *)sb->s_fs_info;
379}
380
381static inline struct ceph_vino ceph_vino(struct inode *inode)
382{
383 return ceph_inode(inode)->i_vino;
384}
385
386/*
387 * ino_t is <64 bits on many architectures, blech.
388 *
389 * i_ino (kernel inode) st_ino (userspace)
390 * i386 32 32
391 * x86_64+ino32 64 32
392 * x86_64 64 64
393 */
394static inline u32 ceph_ino_to_ino32(__u64 vino)
395{
396 u32 ino = vino & 0xffffffff;
397 ino ^= vino >> 32;
398 if (!ino)
399 ino = 2;
400 return ino;
401}
402
403/*
404 * kernel i_ino value
405 */
406static inline ino_t ceph_vino_to_ino(struct ceph_vino vino)
407{
408#if BITS_PER_LONG == 32
409 return ceph_ino_to_ino32(vino.ino);
410#else
411 return (ino_t)vino.ino;
412#endif
413}
414
415/*
416 * user-visible ino (stat, filldir)
417 */
418#if BITS_PER_LONG == 32
419static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino)
420{
421 return ino;
422}
423#else
424static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino)
425{
426 if (ceph_test_mount_opt(ceph_sb_to_client(sb), INO32))
427 ino = ceph_ino_to_ino32(ino);
428 return ino;
429}
430#endif
431
432
433/* for printf-style formatting */
434#define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap
435
436static inline u64 ceph_ino(struct inode *inode)
437{
438 return ceph_inode(inode)->i_vino.ino;
439}
440static inline u64 ceph_snap(struct inode *inode)
441{
442 return ceph_inode(inode)->i_vino.snap;
443}
444
445static inline int ceph_ino_compare(struct inode *inode, void *data)
446{
447 struct ceph_vino *pvino = (struct ceph_vino *)data;
448 struct ceph_inode_info *ci = ceph_inode(inode);
449 return ci->i_vino.ino == pvino->ino &&
450 ci->i_vino.snap == pvino->snap;
451}
452
453static inline struct inode *ceph_find_inode(struct super_block *sb,
454 struct ceph_vino vino)
455{
456 ino_t t = ceph_vino_to_ino(vino);
457 return ilookup5(sb, t, ceph_ino_compare, &vino);
458}
459
460
461/*
462 * Ceph inode.
463 */
464#define CEPH_I_DIR_ORDERED (1 << 0) /* dentries in dir are ordered */
465#define CEPH_I_NODELAY (1 << 1) /* do not delay cap release */
466#define CEPH_I_FLUSH (1 << 2) /* do not delay flush of dirty metadata */
467#define CEPH_I_NOFLUSH (1 << 3) /* do not flush dirty caps */
468#define CEPH_I_POOL_PERM (1 << 4) /* pool rd/wr bits are valid */
469#define CEPH_I_POOL_RD (1 << 5) /* can read from pool */
470#define CEPH_I_POOL_WR (1 << 6) /* can write to pool */
471#define CEPH_I_SEC_INITED (1 << 7) /* security initialized */
472
473static inline void __ceph_dir_set_complete(struct ceph_inode_info *ci,
474 long long release_count,
475 long long ordered_count)
476{
477 smp_mb__before_atomic();
478 atomic64_set(&ci->i_complete_seq[0], release_count);
479 atomic64_set(&ci->i_complete_seq[1], ordered_count);
480}
481
482static inline void __ceph_dir_clear_complete(struct ceph_inode_info *ci)
483{
484 atomic64_inc(&ci->i_release_count);
485}
486
487static inline void __ceph_dir_clear_ordered(struct ceph_inode_info *ci)
488{
489 atomic64_inc(&ci->i_ordered_count);
490}
491
492static inline bool __ceph_dir_is_complete(struct ceph_inode_info *ci)
493{
494 return atomic64_read(&ci->i_complete_seq[0]) ==
495 atomic64_read(&ci->i_release_count);
496}
497
498static inline bool __ceph_dir_is_complete_ordered(struct ceph_inode_info *ci)
499{
500 return atomic64_read(&ci->i_complete_seq[0]) ==
501 atomic64_read(&ci->i_release_count) &&
502 atomic64_read(&ci->i_complete_seq[1]) ==
503 atomic64_read(&ci->i_ordered_count);
504}
505
506static inline void ceph_dir_clear_complete(struct inode *inode)
507{
508 __ceph_dir_clear_complete(ceph_inode(inode));
509}
510
511static inline void ceph_dir_clear_ordered(struct inode *inode)
512{
513 __ceph_dir_clear_ordered(ceph_inode(inode));
514}
515
516static inline bool ceph_dir_is_complete_ordered(struct inode *inode)
517{
518 bool ret = __ceph_dir_is_complete_ordered(ceph_inode(inode));
519 smp_rmb();
520 return ret;
521}
522
523/* find a specific frag @f */
524extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci,
525 u32 f);
526
527/*
528 * choose fragment for value @v. copy frag content to pfrag, if leaf
529 * exists
530 */
531extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v,
532 struct ceph_inode_frag *pfrag,
533 int *found);
534
535static inline struct ceph_dentry_info *ceph_dentry(struct dentry *dentry)
536{
537 return (struct ceph_dentry_info *)dentry->d_fsdata;
538}
539
540static inline loff_t ceph_make_fpos(unsigned frag, unsigned off)
541{
542 return ((loff_t)frag << 32) | (loff_t)off;
543}
544
545/*
546 * caps helpers
547 */
548static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci)
549{
550 return !RB_EMPTY_ROOT(&ci->i_caps);
551}
552
553extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented);
554extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t);
555extern int __ceph_caps_issued_other(struct ceph_inode_info *ci,
556 struct ceph_cap *cap);
557
558static inline int ceph_caps_issued(struct ceph_inode_info *ci)
559{
560 int issued;
561 spin_lock(&ci->i_ceph_lock);
562 issued = __ceph_caps_issued(ci, NULL);
563 spin_unlock(&ci->i_ceph_lock);
564 return issued;
565}
566
567static inline int ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask,
568 int touch)
569{
570 int r;
571 spin_lock(&ci->i_ceph_lock);
572 r = __ceph_caps_issued_mask(ci, mask, touch);
573 spin_unlock(&ci->i_ceph_lock);
574 return r;
575}
576
577static inline int __ceph_caps_dirty(struct ceph_inode_info *ci)
578{
579 return ci->i_dirty_caps | ci->i_flushing_caps;
580}
581extern struct ceph_cap_flush *ceph_alloc_cap_flush(void);
582extern void ceph_free_cap_flush(struct ceph_cap_flush *cf);
583extern int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask,
584 struct ceph_cap_flush **pcf);
585
586extern int __ceph_caps_revoking_other(struct ceph_inode_info *ci,
587 struct ceph_cap *ocap, int mask);
588extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask);
589extern int __ceph_caps_used(struct ceph_inode_info *ci);
590
591extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci);
592
593/*
594 * wanted, by virtue of open file modes AND cap refs (buffered/cached data)
595 */
596static inline int __ceph_caps_wanted(struct ceph_inode_info *ci)
597{
598 int w = __ceph_caps_file_wanted(ci) | __ceph_caps_used(ci);
599 if (w & CEPH_CAP_FILE_BUFFER)
600 w |= CEPH_CAP_FILE_EXCL; /* we want EXCL if dirty data */
601 return w;
602}
603
604/* what the mds thinks we want */
605extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci);
606
607extern void ceph_caps_init(struct ceph_mds_client *mdsc);
608extern void ceph_caps_finalize(struct ceph_mds_client *mdsc);
609extern void ceph_adjust_min_caps(struct ceph_mds_client *mdsc, int delta);
610extern void ceph_reserve_caps(struct ceph_mds_client *mdsc,
611 struct ceph_cap_reservation *ctx, int need);
612extern int ceph_unreserve_caps(struct ceph_mds_client *mdsc,
613 struct ceph_cap_reservation *ctx);
614extern void ceph_reservation_status(struct ceph_fs_client *client,
615 int *total, int *avail, int *used,
616 int *reserved, int *min);
617
618
619
620/*
621 * we keep buffered readdir results attached to file->private_data
622 */
623#define CEPH_F_SYNC 1
624#define CEPH_F_ATEND 2
625
626struct ceph_file_info {
627 short fmode; /* initialized on open */
628 short flags; /* CEPH_F_* */
629
630 /* readdir: position within the dir */
631 u32 frag;
632 struct ceph_mds_request *last_readdir;
633
634 /* readdir: position within a frag */
635 unsigned offset; /* offset of last chunk, adjusted for . and .. */
636 unsigned next_offset; /* offset of next chunk (last_name's + 1) */
637 char *last_name; /* last entry in previous chunk */
638 long long dir_release_count;
639 long long dir_ordered_count;
640 int readdir_cache_idx;
641
642 /* used for -o dirstat read() on directory thing */
643 char *dir_info;
644 int dir_info_len;
645};
646
647struct ceph_readdir_cache_control {
648 struct page *page;
649 struct dentry **dentries;
650 int index;
651};
652
653/*
654 * A "snap realm" describes a subset of the file hierarchy sharing
655 * the same set of snapshots that apply to it. The realms themselves
656 * are organized into a hierarchy, such that children inherit (some of)
657 * the snapshots of their parents.
658 *
659 * All inodes within the realm that have capabilities are linked into a
660 * per-realm list.
661 */
662struct ceph_snap_realm {
663 u64 ino;
664 atomic_t nref;
665 struct rb_node node;
666
667 u64 created, seq;
668 u64 parent_ino;
669 u64 parent_since; /* snapid when our current parent became so */
670
671 u64 *prior_parent_snaps; /* snaps inherited from any parents we */
672 u32 num_prior_parent_snaps; /* had prior to parent_since */
673 u64 *snaps; /* snaps specific to this realm */
674 u32 num_snaps;
675
676 struct ceph_snap_realm *parent;
677 struct list_head children; /* list of child realms */
678 struct list_head child_item;
679
680 struct list_head empty_item; /* if i have ref==0 */
681
682 struct list_head dirty_item; /* if realm needs new context */
683
684 /* the current set of snaps for this realm */
685 struct ceph_snap_context *cached_context;
686
687 struct list_head inodes_with_caps;
688 spinlock_t inodes_with_caps_lock;
689};
690
691static inline int default_congestion_kb(void)
692{
693 int congestion_kb;
694
695 /*
696 * Copied from NFS
697 *
698 * congestion size, scale with available memory.
699 *
700 * 64MB: 8192k
701 * 128MB: 11585k
702 * 256MB: 16384k
703 * 512MB: 23170k
704 * 1GB: 32768k
705 * 2GB: 46340k
706 * 4GB: 65536k
707 * 8GB: 92681k
708 * 16GB: 131072k
709 *
710 * This allows larger machines to have larger/more transfers.
711 * Limit the default to 256M
712 */
713 congestion_kb = (16*int_sqrt(totalram_pages)) << (PAGE_SHIFT-10);
714 if (congestion_kb > 256*1024)
715 congestion_kb = 256*1024;
716
717 return congestion_kb;
718}
719
720
721
722/* snap.c */
723struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc,
724 u64 ino);
725extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc,
726 struct ceph_snap_realm *realm);
727extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc,
728 struct ceph_snap_realm *realm);
729extern int ceph_update_snap_trace(struct ceph_mds_client *m,
730 void *p, void *e, bool deletion,
731 struct ceph_snap_realm **realm_ret);
732extern void ceph_handle_snap(struct ceph_mds_client *mdsc,
733 struct ceph_mds_session *session,
734 struct ceph_msg *msg);
735extern void ceph_queue_cap_snap(struct ceph_inode_info *ci);
736extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci,
737 struct ceph_cap_snap *capsnap);
738extern void ceph_cleanup_empty_realms(struct ceph_mds_client *mdsc);
739
740/*
741 * a cap_snap is "pending" if it is still awaiting an in-progress
742 * sync write (that may/may not still update size, mtime, etc.).
743 */
744static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci)
745{
746 return !list_empty(&ci->i_cap_snaps) &&
747 list_last_entry(&ci->i_cap_snaps, struct ceph_cap_snap,
748 ci_item)->writing;
749}
750
751/* inode.c */
752extern const struct inode_operations ceph_file_iops;
753
754extern struct inode *ceph_alloc_inode(struct super_block *sb);
755extern void ceph_destroy_inode(struct inode *inode);
756extern int ceph_drop_inode(struct inode *inode);
757
758extern struct inode *ceph_get_inode(struct super_block *sb,
759 struct ceph_vino vino);
760extern struct inode *ceph_get_snapdir(struct inode *parent);
761extern int ceph_fill_file_size(struct inode *inode, int issued,
762 u32 truncate_seq, u64 truncate_size, u64 size);
763extern void ceph_fill_file_time(struct inode *inode, int issued,
764 u64 time_warp_seq, struct timespec *ctime,
765 struct timespec *mtime, struct timespec *atime);
766extern int ceph_fill_trace(struct super_block *sb,
767 struct ceph_mds_request *req,
768 struct ceph_mds_session *session);
769extern int ceph_readdir_prepopulate(struct ceph_mds_request *req,
770 struct ceph_mds_session *session);
771
772extern int ceph_inode_holds_cap(struct inode *inode, int mask);
773
774extern int ceph_inode_set_size(struct inode *inode, loff_t size);
775extern void __ceph_do_pending_vmtruncate(struct inode *inode);
776extern void ceph_queue_vmtruncate(struct inode *inode);
777
778extern void ceph_queue_invalidate(struct inode *inode);
779extern void ceph_queue_writeback(struct inode *inode);
780
781extern int __ceph_do_getattr(struct inode *inode, struct page *locked_page,
782 int mask, bool force);
783static inline int ceph_do_getattr(struct inode *inode, int mask, bool force)
784{
785 return __ceph_do_getattr(inode, NULL, mask, force);
786}
787extern int ceph_permission(struct inode *inode, int mask);
788extern int ceph_setattr(struct dentry *dentry, struct iattr *attr);
789extern int ceph_getattr(struct vfsmount *mnt, struct dentry *dentry,
790 struct kstat *stat);
791
792/* xattr.c */
793extern int ceph_setxattr(struct dentry *, const char *, const void *,
794 size_t, int);
795int __ceph_setxattr(struct dentry *, const char *, const void *, size_t, int);
796ssize_t __ceph_getxattr(struct inode *, const char *, void *, size_t);
797int __ceph_removexattr(struct dentry *, const char *);
798extern ssize_t ceph_getxattr(struct dentry *, const char *, void *, size_t);
799extern ssize_t ceph_listxattr(struct dentry *, char *, size_t);
800extern int ceph_removexattr(struct dentry *, const char *);
801extern void __ceph_build_xattrs_blob(struct ceph_inode_info *ci);
802extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci);
803extern void __init ceph_xattr_init(void);
804extern void ceph_xattr_exit(void);
805extern const struct xattr_handler *ceph_xattr_handlers[];
806
807#ifdef CONFIG_SECURITY
808extern bool ceph_security_xattr_deadlock(struct inode *in);
809extern bool ceph_security_xattr_wanted(struct inode *in);
810#else
811static inline bool ceph_security_xattr_deadlock(struct inode *in)
812{
813 return false;
814}
815static inline bool ceph_security_xattr_wanted(struct inode *in)
816{
817 return false;
818}
819#endif
820
821/* acl.c */
822struct ceph_acls_info {
823 void *default_acl;
824 void *acl;
825 struct ceph_pagelist *pagelist;
826};
827
828#ifdef CONFIG_CEPH_FS_POSIX_ACL
829
830struct posix_acl *ceph_get_acl(struct inode *, int);
831int ceph_set_acl(struct inode *inode, struct posix_acl *acl, int type);
832int ceph_pre_init_acls(struct inode *dir, umode_t *mode,
833 struct ceph_acls_info *info);
834void ceph_init_inode_acls(struct inode *inode, struct ceph_acls_info *info);
835void ceph_release_acls_info(struct ceph_acls_info *info);
836
837static inline void ceph_forget_all_cached_acls(struct inode *inode)
838{
839 forget_all_cached_acls(inode);
840}
841
842#else
843
844#define ceph_get_acl NULL
845#define ceph_set_acl NULL
846
847static inline int ceph_pre_init_acls(struct inode *dir, umode_t *mode,
848 struct ceph_acls_info *info)
849{
850 return 0;
851}
852static inline void ceph_init_inode_acls(struct inode *inode,
853 struct ceph_acls_info *info)
854{
855}
856static inline void ceph_release_acls_info(struct ceph_acls_info *info)
857{
858}
859static inline int ceph_acl_chmod(struct dentry *dentry, struct inode *inode)
860{
861 return 0;
862}
863
864static inline void ceph_forget_all_cached_acls(struct inode *inode)
865{
866}
867
868#endif
869
870/* caps.c */
871extern const char *ceph_cap_string(int c);
872extern void ceph_handle_caps(struct ceph_mds_session *session,
873 struct ceph_msg *msg);
874extern struct ceph_cap *ceph_get_cap(struct ceph_mds_client *mdsc,
875 struct ceph_cap_reservation *ctx);
876extern void ceph_add_cap(struct inode *inode,
877 struct ceph_mds_session *session, u64 cap_id,
878 int fmode, unsigned issued, unsigned wanted,
879 unsigned cap, unsigned seq, u64 realmino, int flags,
880 struct ceph_cap **new_cap);
881extern void __ceph_remove_cap(struct ceph_cap *cap, bool queue_release);
882extern void ceph_put_cap(struct ceph_mds_client *mdsc,
883 struct ceph_cap *cap);
884extern int ceph_is_any_caps(struct inode *inode);
885
886extern void ceph_queue_caps_release(struct inode *inode);
887extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc);
888extern int ceph_fsync(struct file *file, loff_t start, loff_t end,
889 int datasync);
890extern void ceph_early_kick_flushing_caps(struct ceph_mds_client *mdsc,
891 struct ceph_mds_session *session);
892extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc,
893 struct ceph_mds_session *session);
894extern struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci,
895 int mds);
896extern int ceph_get_cap_mds(struct inode *inode);
897extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps);
898extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had);
899extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
900 struct ceph_snap_context *snapc);
901extern void __ceph_flush_snaps(struct ceph_inode_info *ci,
902 struct ceph_mds_session **psession,
903 int again);
904extern void ceph_check_caps(struct ceph_inode_info *ci, int flags,
905 struct ceph_mds_session *session);
906extern void ceph_check_delayed_caps(struct ceph_mds_client *mdsc);
907extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc);
908
909extern int ceph_encode_inode_release(void **p, struct inode *inode,
910 int mds, int drop, int unless, int force);
911extern int ceph_encode_dentry_release(void **p, struct dentry *dn,
912 int mds, int drop, int unless);
913
914extern int ceph_get_caps(struct ceph_inode_info *ci, int need, int want,
915 loff_t endoff, int *got, struct page **pinned_page);
916
917/* for counting open files by mode */
918static inline void __ceph_get_fmode(struct ceph_inode_info *ci, int mode)
919{
920 ci->i_nr_by_mode[mode]++;
921}
922extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode);
923
924/* addr.c */
925extern const struct address_space_operations ceph_aops;
926extern int ceph_mmap(struct file *file, struct vm_area_struct *vma);
927extern int ceph_uninline_data(struct file *filp, struct page *locked_page);
928extern int ceph_pool_perm_check(struct ceph_inode_info *ci, int need);
929extern void ceph_pool_perm_destroy(struct ceph_mds_client* mdsc);
930
931/* file.c */
932extern const struct file_operations ceph_file_fops;
933
934extern int ceph_open(struct inode *inode, struct file *file);
935extern int ceph_atomic_open(struct inode *dir, struct dentry *dentry,
936 struct file *file, unsigned flags, umode_t mode,
937 int *opened);
938extern int ceph_release(struct inode *inode, struct file *filp);
939extern void ceph_fill_inline_data(struct inode *inode, struct page *locked_page,
940 char *data, size_t len);
941/* dir.c */
942extern const struct file_operations ceph_dir_fops;
943extern const struct file_operations ceph_snapdir_fops;
944extern const struct inode_operations ceph_dir_iops;
945extern const struct inode_operations ceph_snapdir_iops;
946extern const struct dentry_operations ceph_dentry_ops, ceph_snap_dentry_ops,
947 ceph_snapdir_dentry_ops;
948
949extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry);
950extern int ceph_handle_snapdir(struct ceph_mds_request *req,
951 struct dentry *dentry, int err);
952extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req,
953 struct dentry *dentry, int err);
954
955extern void ceph_dentry_lru_add(struct dentry *dn);
956extern void ceph_dentry_lru_touch(struct dentry *dn);
957extern void ceph_dentry_lru_del(struct dentry *dn);
958extern void ceph_invalidate_dentry_lease(struct dentry *dentry);
959extern unsigned ceph_dentry_hash(struct inode *dir, struct dentry *dn);
960extern void ceph_readdir_cache_release(struct ceph_readdir_cache_control *ctl);
961
962/*
963 * our d_ops vary depending on whether the inode is live,
964 * snapshotted (read-only), or a virtual ".snap" directory.
965 */
966int ceph_init_dentry(struct dentry *dentry);
967
968
969/* ioctl.c */
970extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
971
972/* export.c */
973extern const struct export_operations ceph_export_ops;
974
975/* locks.c */
976extern __init void ceph_flock_init(void);
977extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl);
978extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl);
979extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num);
980extern int ceph_encode_locks_to_buffer(struct inode *inode,
981 struct ceph_filelock *flocks,
982 int num_fcntl_locks,
983 int num_flock_locks);
984extern int ceph_locks_to_pagelist(struct ceph_filelock *flocks,
985 struct ceph_pagelist *pagelist,
986 int num_fcntl_locks, int num_flock_locks);
987extern int lock_to_ceph_filelock(struct file_lock *fl, struct ceph_filelock *c);
988
989/* debugfs.c */
990extern int ceph_fs_debugfs_init(struct ceph_fs_client *client);
991extern void ceph_fs_debugfs_cleanup(struct ceph_fs_client *client);
992
993#endif /* _FS_CEPH_SUPER_H */
1/* SPDX-License-Identifier: GPL-2.0 */
2#ifndef _FS_CEPH_SUPER_H
3#define _FS_CEPH_SUPER_H
4
5#include <linux/ceph/ceph_debug.h>
6
7#include <asm/unaligned.h>
8#include <linux/backing-dev.h>
9#include <linux/completion.h>
10#include <linux/exportfs.h>
11#include <linux/fs.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>
20
21#include <linux/ceph/libceph.h>
22
23#ifdef CONFIG_CEPH_FSCACHE
24#include <linux/fscache.h>
25#endif
26
27/* f_type in struct statfs */
28#define CEPH_SUPER_MAGIC 0x00c36400
29
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)
34
35#define CEPH_MOUNT_OPT_CLEANRECOVER (1<<1) /* auto reonnect (clean mode) after blacklisted */
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 */
47
48#define CEPH_MOUNT_OPT_DEFAULT \
49 (CEPH_MOUNT_OPT_DCACHE | \
50 CEPH_MOUNT_OPT_NOCOPYFROM)
51
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))
58
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"
68
69/*
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.
74 */
75#define CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT 5 /* cap release delay */
76#define CEPH_CAPS_WANTED_DELAY_MAX_DEFAULT 60 /* cap release delay */
77
78struct ceph_mount_options {
79 unsigned int flags;
80
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;
86 int caps_max;
87 unsigned int max_readdir; /* max readdir result (entries) */
88 unsigned int max_readdir_bytes; /* max readdir result (bytes) */
89
90 /*
91 * everything above this point can be memcmp'd; everything below
92 * is handled in compare_mount_options()
93 */
94
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 */
99};
100
101struct ceph_fs_client {
102 struct super_block *sb;
103
104 struct list_head metric_wakeup;
105
106 struct ceph_mount_options *mount_options;
107 struct ceph_client *client;
108
109 unsigned long mount_state;
110
111 unsigned long last_auto_reconnect;
112 bool blacklisted;
113
114 bool have_copy_from2;
115
116 u32 filp_gen;
117 loff_t max_file_size;
118
119 struct ceph_mds_client *mdsc;
120
121 atomic_long_t writeback_count;
122
123 struct workqueue_struct *inode_wq;
124 struct workqueue_struct *cap_wq;
125
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;
133#endif
134
135#ifdef CONFIG_CEPH_FSCACHE
136 struct fscache_cookie *fscache;
137#endif
138};
139
140
141/*
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.
147 *
148 * Each cap is referenced by the inode's i_caps rbtree and by per-mds
149 * session capability lists.
150 */
151struct ceph_cap {
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) */
157 union {
158 /* in-use caps */
159 struct {
160 int issued; /* latest, from the mds */
161 int implemented; /* implemented superset of
162 issued (for revocation) */
163 int mds, mds_wanted;
164 };
165 /* caps to release */
166 struct {
167 u64 cap_ino;
168 int queue_release;
169 };
170 };
171 u32 seq, issue_seq, mseq;
172 u32 cap_gen; /* active/stale cycle */
173 unsigned long last_used;
174 struct list_head caps_item;
175};
176
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 */
180
181struct ceph_cap_flush {
182 u64 tid;
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
187};
188
189/*
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.
193 */
194struct ceph_cap_snap {
195 refcount_t nref;
196 struct list_head ci_item;
197
198 struct ceph_cap_flush cap_flush;
199
200 u64 follows;
201 int issued, dirty;
202 struct ceph_snap_context *context;
203
204 umode_t mode;
205 kuid_t uid;
206 kgid_t gid;
207
208 struct ceph_buffer *xattr_blob;
209 u64 xattr_version;
210
211 u64 size;
212 u64 change_attr;
213 struct timespec64 mtime, atime, ctime, btime;
214 u64 time_warp_seq;
215 u64 truncate_size;
216 u32 truncate_seq;
217 int writing; /* a sync write is still in progress */
218 int dirty_pages; /* dirty pages awaiting writeback */
219 bool inline_data;
220 bool need_flush;
221};
222
223static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap)
224{
225 if (refcount_dec_and_test(&capsnap->nref)) {
226 if (capsnap->xattr_blob)
227 ceph_buffer_put(capsnap->xattr_blob);
228 kfree(capsnap);
229 }
230}
231
232/*
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.
236 *
237 * A _leaf_ frag will be present in the i_fragtree IFF there is
238 * delegation info. That is, if mds >= 0 || ndist > 0.
239 */
240#define CEPH_MAX_DIRFRAG_REP 4
241
242struct ceph_inode_frag {
243 struct rb_node node;
244
245 /* fragtree state */
246 u32 frag;
247 int split_by; /* i.e. 2^(split_by) children */
248
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];
253};
254
255/*
256 * We cache inode xattrs as an encoded blob until they are first used,
257 * at which point we parse them into an rbtree.
258 */
259struct ceph_inode_xattr {
260 struct rb_node node;
261
262 const char *name;
263 int name_len;
264 const char *val;
265 int val_len;
266 int dirty;
267
268 int should_free_name;
269 int should_free_val;
270};
271
272/*
273 * Ceph dentry state
274 */
275struct ceph_dentry_info {
276 struct dentry *dentry;
277 struct ceph_mds_session *lease_session;
278 struct list_head lease_list;
279 unsigned flags;
280 int lease_shared_gen;
281 u32 lease_gen;
282 u32 lease_seq;
283 unsigned long lease_renew_after, lease_renew_from;
284 unsigned long time;
285 u64 offset;
286};
287
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
292
293struct ceph_inode_xattrs_info {
294 /*
295 * (still encoded) xattr blob. we avoid the overhead of parsing
296 * this until someone actually calls getxattr, etc.
297 *
298 * blob->vec.iov_len == 4 implies there are no xattrs; blob ==
299 * NULL means we don't know.
300 */
301 struct ceph_buffer *blob, *prealloc_blob;
302
303 struct rb_root index;
304 bool dirty;
305 int count;
306 int names_size;
307 int vals_size;
308 u64 version, index_version;
309};
310
311/*
312 * Ceph inode.
313 */
314struct ceph_inode_info {
315 struct ceph_vino i_vino; /* ceph ino + snap */
316
317 spinlock_t i_ceph_lock;
318
319 u64 i_version;
320 u64 i_inline_version;
321 u32 i_time_warp_seq;
322
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];
327
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
331 char *i_symlink;
332
333 /* for dirs */
334 struct timespec64 i_rctime;
335 u64 i_rbytes, i_rfiles, i_rsubdirs;
336 u64 i_files, i_subdirs;
337
338 /* quotas */
339 u64 i_max_bytes, i_max_files;
340
341 s32 i_dir_pin;
342
343 struct rb_root i_fragtree;
344 int i_fragtree_nsplits;
345 struct mutex i_fragtree_mutex;
346
347 struct ceph_inode_xattrs_info i_xattrs;
348
349 /* capabilities. protected _both_ by i_ceph_lock and cap->session's
350 * s_mutex. */
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 */
354
355 /*
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.
361 */
362 struct list_head i_dirty_item;
363
364 /*
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.
370 */
371 struct list_head i_flushing_item;
372
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 */
386
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 */
390
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 */
395
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 */
400
401 /* held references to caps */
402 int i_pin_ref;
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 */
409
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;
413
414 union {
415 struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */
416 struct ceph_snapid_map *i_snapid_map; /* snapid -> dev_t */
417 };
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;
423
424 struct work_struct i_work;
425 unsigned long i_work_mask;
426
427#ifdef CONFIG_CEPH_FSCACHE
428 struct fscache_cookie *fscache;
429 u32 i_fscache_gen;
430#endif
431 errseq_t i_meta_err;
432
433 struct inode vfs_inode; /* at end */
434};
435
436static inline struct ceph_inode_info *
437ceph_inode(const struct inode *inode)
438{
439 return container_of(inode, struct ceph_inode_info, vfs_inode);
440}
441
442static inline struct ceph_fs_client *
443ceph_inode_to_client(const struct inode *inode)
444{
445 return (struct ceph_fs_client *)inode->i_sb->s_fs_info;
446}
447
448static inline struct ceph_fs_client *
449ceph_sb_to_client(const struct super_block *sb)
450{
451 return (struct ceph_fs_client *)sb->s_fs_info;
452}
453
454static inline struct ceph_vino
455ceph_vino(const struct inode *inode)
456{
457 return ceph_inode(inode)->i_vino;
458}
459
460static inline u32 ceph_ino_to_ino32(u64 vino)
461{
462 u32 ino = vino & 0xffffffff;
463 ino ^= vino >> 32;
464 if (!ino)
465 ino = 2;
466 return ino;
467}
468
469/*
470 * Inode numbers in cephfs are 64 bits, but inode->i_ino is 32-bits on
471 * some arches. We generally do not use this value inside the ceph driver, but
472 * we do want to set it to something, so that generic vfs code has an
473 * appropriate value for tracepoints and the like.
474 */
475static inline ino_t ceph_vino_to_ino_t(struct ceph_vino vino)
476{
477 if (sizeof(ino_t) == sizeof(u32))
478 return ceph_ino_to_ino32(vino.ino);
479 return (ino_t)vino.ino;
480}
481
482/* for printf-style formatting */
483#define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap
484
485static inline u64 ceph_ino(struct inode *inode)
486{
487 return ceph_inode(inode)->i_vino.ino;
488}
489
490static inline u64 ceph_snap(struct inode *inode)
491{
492 return ceph_inode(inode)->i_vino.snap;
493}
494
495/**
496 * ceph_present_ino - format an inode number for presentation to userland
497 * @sb: superblock where the inode lives
498 * @ino: inode number to (possibly) convert
499 *
500 * If the user mounted with the ino32 option, then the 64-bit value needs
501 * to be converted to something that can fit inside 32 bits. Note that
502 * internal kernel code never uses this value, so this is entirely for
503 * userland consumption.
504 */
505static inline u64 ceph_present_ino(struct super_block *sb, u64 ino)
506{
507 if (unlikely(ceph_test_mount_opt(ceph_sb_to_client(sb), INO32)))
508 return ceph_ino_to_ino32(ino);
509 return ino;
510}
511
512static inline u64 ceph_present_inode(struct inode *inode)
513{
514 return ceph_present_ino(inode->i_sb, ceph_ino(inode));
515}
516
517static inline int ceph_ino_compare(struct inode *inode, void *data)
518{
519 struct ceph_vino *pvino = (struct ceph_vino *)data;
520 struct ceph_inode_info *ci = ceph_inode(inode);
521 return ci->i_vino.ino == pvino->ino &&
522 ci->i_vino.snap == pvino->snap;
523}
524
525
526static inline struct inode *ceph_find_inode(struct super_block *sb,
527 struct ceph_vino vino)
528{
529 /*
530 * NB: The hashval will be run through the fs/inode.c hash function
531 * anyway, so there is no need to squash the inode number down to
532 * 32-bits first. Just use low-order bits on arches with 32-bit long.
533 */
534 return ilookup5(sb, (unsigned long)vino.ino, ceph_ino_compare, &vino);
535}
536
537
538/*
539 * Ceph inode.
540 */
541#define CEPH_I_DIR_ORDERED (1 << 0) /* dentries in dir are ordered */
542#define CEPH_I_FLUSH (1 << 2) /* do not delay flush of dirty metadata */
543#define CEPH_I_POOL_PERM (1 << 3) /* pool rd/wr bits are valid */
544#define CEPH_I_POOL_RD (1 << 4) /* can read from pool */
545#define CEPH_I_POOL_WR (1 << 5) /* can write to pool */
546#define CEPH_I_SEC_INITED (1 << 6) /* security initialized */
547#define CEPH_I_KICK_FLUSH (1 << 7) /* kick flushing caps */
548#define CEPH_I_FLUSH_SNAPS (1 << 8) /* need flush snapss */
549#define CEPH_I_ERROR_WRITE (1 << 9) /* have seen write errors */
550#define CEPH_I_ERROR_FILELOCK (1 << 10) /* have seen file lock errors */
551#define CEPH_I_ODIRECT (1 << 11) /* inode in direct I/O mode */
552#define CEPH_ASYNC_CREATE_BIT (12) /* async create in flight for this */
553#define CEPH_I_ASYNC_CREATE (1 << CEPH_ASYNC_CREATE_BIT)
554
555/*
556 * Masks of ceph inode work.
557 */
558#define CEPH_I_WORK_WRITEBACK 0 /* writeback */
559#define CEPH_I_WORK_INVALIDATE_PAGES 1 /* invalidate pages */
560#define CEPH_I_WORK_VMTRUNCATE 2 /* vmtruncate */
561
562/*
563 * We set the ERROR_WRITE bit when we start seeing write errors on an inode
564 * and then clear it when they start succeeding. Note that we do a lockless
565 * check first, and only take the lock if it looks like it needs to be changed.
566 * The write submission code just takes this as a hint, so we're not too
567 * worried if a few slip through in either direction.
568 */
569static inline void ceph_set_error_write(struct ceph_inode_info *ci)
570{
571 if (!(READ_ONCE(ci->i_ceph_flags) & CEPH_I_ERROR_WRITE)) {
572 spin_lock(&ci->i_ceph_lock);
573 ci->i_ceph_flags |= CEPH_I_ERROR_WRITE;
574 spin_unlock(&ci->i_ceph_lock);
575 }
576}
577
578static inline void ceph_clear_error_write(struct ceph_inode_info *ci)
579{
580 if (READ_ONCE(ci->i_ceph_flags) & CEPH_I_ERROR_WRITE) {
581 spin_lock(&ci->i_ceph_lock);
582 ci->i_ceph_flags &= ~CEPH_I_ERROR_WRITE;
583 spin_unlock(&ci->i_ceph_lock);
584 }
585}
586
587static inline void __ceph_dir_set_complete(struct ceph_inode_info *ci,
588 long long release_count,
589 long long ordered_count)
590{
591 /*
592 * Makes sure operations that setup readdir cache (update page
593 * cache and i_size) are strongly ordered w.r.t. the following
594 * atomic64_set() operations.
595 */
596 smp_mb();
597 atomic64_set(&ci->i_complete_seq[0], release_count);
598 atomic64_set(&ci->i_complete_seq[1], ordered_count);
599}
600
601static inline void __ceph_dir_clear_complete(struct ceph_inode_info *ci)
602{
603 atomic64_inc(&ci->i_release_count);
604}
605
606static inline void __ceph_dir_clear_ordered(struct ceph_inode_info *ci)
607{
608 atomic64_inc(&ci->i_ordered_count);
609}
610
611static inline bool __ceph_dir_is_complete(struct ceph_inode_info *ci)
612{
613 return atomic64_read(&ci->i_complete_seq[0]) ==
614 atomic64_read(&ci->i_release_count);
615}
616
617static inline bool __ceph_dir_is_complete_ordered(struct ceph_inode_info *ci)
618{
619 return atomic64_read(&ci->i_complete_seq[0]) ==
620 atomic64_read(&ci->i_release_count) &&
621 atomic64_read(&ci->i_complete_seq[1]) ==
622 atomic64_read(&ci->i_ordered_count);
623}
624
625static inline void ceph_dir_clear_complete(struct inode *inode)
626{
627 __ceph_dir_clear_complete(ceph_inode(inode));
628}
629
630static inline void ceph_dir_clear_ordered(struct inode *inode)
631{
632 __ceph_dir_clear_ordered(ceph_inode(inode));
633}
634
635static inline bool ceph_dir_is_complete_ordered(struct inode *inode)
636{
637 bool ret = __ceph_dir_is_complete_ordered(ceph_inode(inode));
638 smp_rmb();
639 return ret;
640}
641
642/* find a specific frag @f */
643extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci,
644 u32 f);
645
646/*
647 * choose fragment for value @v. copy frag content to pfrag, if leaf
648 * exists
649 */
650extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v,
651 struct ceph_inode_frag *pfrag,
652 int *found);
653
654static inline struct ceph_dentry_info *ceph_dentry(const struct dentry *dentry)
655{
656 return (struct ceph_dentry_info *)dentry->d_fsdata;
657}
658
659/*
660 * caps helpers
661 */
662static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci)
663{
664 return !RB_EMPTY_ROOT(&ci->i_caps);
665}
666
667extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented);
668extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t);
669extern int __ceph_caps_issued_mask_metric(struct ceph_inode_info *ci, int mask,
670 int t);
671extern int __ceph_caps_issued_other(struct ceph_inode_info *ci,
672 struct ceph_cap *cap);
673
674static inline int ceph_caps_issued(struct ceph_inode_info *ci)
675{
676 int issued;
677 spin_lock(&ci->i_ceph_lock);
678 issued = __ceph_caps_issued(ci, NULL);
679 spin_unlock(&ci->i_ceph_lock);
680 return issued;
681}
682
683static inline int ceph_caps_issued_mask_metric(struct ceph_inode_info *ci,
684 int mask, int touch)
685{
686 int r;
687 spin_lock(&ci->i_ceph_lock);
688 r = __ceph_caps_issued_mask_metric(ci, mask, touch);
689 spin_unlock(&ci->i_ceph_lock);
690 return r;
691}
692
693static inline int __ceph_caps_dirty(struct ceph_inode_info *ci)
694{
695 return ci->i_dirty_caps | ci->i_flushing_caps;
696}
697extern struct ceph_cap_flush *ceph_alloc_cap_flush(void);
698extern void ceph_free_cap_flush(struct ceph_cap_flush *cf);
699extern int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask,
700 struct ceph_cap_flush **pcf);
701
702extern int __ceph_caps_revoking_other(struct ceph_inode_info *ci,
703 struct ceph_cap *ocap, int mask);
704extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask);
705extern int __ceph_caps_used(struct ceph_inode_info *ci);
706
707static inline bool __ceph_is_file_opened(struct ceph_inode_info *ci)
708{
709 return ci->i_nr_by_mode[0];
710}
711extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci);
712extern int __ceph_caps_wanted(struct ceph_inode_info *ci);
713
714/* what the mds thinks we want */
715extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci, bool check);
716
717extern void ceph_caps_init(struct ceph_mds_client *mdsc);
718extern void ceph_caps_finalize(struct ceph_mds_client *mdsc);
719extern void ceph_adjust_caps_max_min(struct ceph_mds_client *mdsc,
720 struct ceph_mount_options *fsopt);
721extern int ceph_reserve_caps(struct ceph_mds_client *mdsc,
722 struct ceph_cap_reservation *ctx, int need);
723extern void ceph_unreserve_caps(struct ceph_mds_client *mdsc,
724 struct ceph_cap_reservation *ctx);
725extern void ceph_reservation_status(struct ceph_fs_client *client,
726 int *total, int *avail, int *used,
727 int *reserved, int *min);
728
729
730
731/*
732 * we keep buffered readdir results attached to file->private_data
733 */
734#define CEPH_F_SYNC 1
735#define CEPH_F_ATEND 2
736
737struct ceph_file_info {
738 short fmode; /* initialized on open */
739 short flags; /* CEPH_F_* */
740
741 spinlock_t rw_contexts_lock;
742 struct list_head rw_contexts;
743
744 errseq_t meta_err;
745 u32 filp_gen;
746 atomic_t num_locks;
747};
748
749struct ceph_dir_file_info {
750 struct ceph_file_info file_info;
751
752 /* readdir: position within the dir */
753 u32 frag;
754 struct ceph_mds_request *last_readdir;
755
756 /* readdir: position within a frag */
757 unsigned next_offset; /* offset of next chunk (last_name's + 1) */
758 char *last_name; /* last entry in previous chunk */
759 long long dir_release_count;
760 long long dir_ordered_count;
761 int readdir_cache_idx;
762
763 /* used for -o dirstat read() on directory thing */
764 char *dir_info;
765 int dir_info_len;
766};
767
768struct ceph_rw_context {
769 struct list_head list;
770 struct task_struct *thread;
771 int caps;
772};
773
774#define CEPH_DEFINE_RW_CONTEXT(_name, _caps) \
775 struct ceph_rw_context _name = { \
776 .thread = current, \
777 .caps = _caps, \
778 }
779
780static inline void ceph_add_rw_context(struct ceph_file_info *cf,
781 struct ceph_rw_context *ctx)
782{
783 spin_lock(&cf->rw_contexts_lock);
784 list_add(&ctx->list, &cf->rw_contexts);
785 spin_unlock(&cf->rw_contexts_lock);
786}
787
788static inline void ceph_del_rw_context(struct ceph_file_info *cf,
789 struct ceph_rw_context *ctx)
790{
791 spin_lock(&cf->rw_contexts_lock);
792 list_del(&ctx->list);
793 spin_unlock(&cf->rw_contexts_lock);
794}
795
796static inline struct ceph_rw_context*
797ceph_find_rw_context(struct ceph_file_info *cf)
798{
799 struct ceph_rw_context *ctx, *found = NULL;
800 spin_lock(&cf->rw_contexts_lock);
801 list_for_each_entry(ctx, &cf->rw_contexts, list) {
802 if (ctx->thread == current) {
803 found = ctx;
804 break;
805 }
806 }
807 spin_unlock(&cf->rw_contexts_lock);
808 return found;
809}
810
811struct ceph_readdir_cache_control {
812 struct page *page;
813 struct dentry **dentries;
814 int index;
815};
816
817/*
818 * A "snap realm" describes a subset of the file hierarchy sharing
819 * the same set of snapshots that apply to it. The realms themselves
820 * are organized into a hierarchy, such that children inherit (some of)
821 * the snapshots of their parents.
822 *
823 * All inodes within the realm that have capabilities are linked into a
824 * per-realm list.
825 */
826struct ceph_snap_realm {
827 u64 ino;
828 struct inode *inode;
829 atomic_t nref;
830 struct rb_node node;
831
832 u64 created, seq;
833 u64 parent_ino;
834 u64 parent_since; /* snapid when our current parent became so */
835
836 u64 *prior_parent_snaps; /* snaps inherited from any parents we */
837 u32 num_prior_parent_snaps; /* had prior to parent_since */
838 u64 *snaps; /* snaps specific to this realm */
839 u32 num_snaps;
840
841 struct ceph_snap_realm *parent;
842 struct list_head children; /* list of child realms */
843 struct list_head child_item;
844
845 struct list_head empty_item; /* if i have ref==0 */
846
847 struct list_head dirty_item; /* if realm needs new context */
848
849 /* the current set of snaps for this realm */
850 struct ceph_snap_context *cached_context;
851
852 struct list_head inodes_with_caps;
853 spinlock_t inodes_with_caps_lock;
854};
855
856static inline int default_congestion_kb(void)
857{
858 int congestion_kb;
859
860 /*
861 * Copied from NFS
862 *
863 * congestion size, scale with available memory.
864 *
865 * 64MB: 8192k
866 * 128MB: 11585k
867 * 256MB: 16384k
868 * 512MB: 23170k
869 * 1GB: 32768k
870 * 2GB: 46340k
871 * 4GB: 65536k
872 * 8GB: 92681k
873 * 16GB: 131072k
874 *
875 * This allows larger machines to have larger/more transfers.
876 * Limit the default to 256M
877 */
878 congestion_kb = (16*int_sqrt(totalram_pages())) << (PAGE_SHIFT-10);
879 if (congestion_kb > 256*1024)
880 congestion_kb = 256*1024;
881
882 return congestion_kb;
883}
884
885
886/* super.c */
887extern int ceph_force_reconnect(struct super_block *sb);
888/* snap.c */
889struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc,
890 u64 ino);
891extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc,
892 struct ceph_snap_realm *realm);
893extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc,
894 struct ceph_snap_realm *realm);
895extern int ceph_update_snap_trace(struct ceph_mds_client *m,
896 void *p, void *e, bool deletion,
897 struct ceph_snap_realm **realm_ret);
898extern void ceph_handle_snap(struct ceph_mds_client *mdsc,
899 struct ceph_mds_session *session,
900 struct ceph_msg *msg);
901extern void ceph_queue_cap_snap(struct ceph_inode_info *ci);
902extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci,
903 struct ceph_cap_snap *capsnap);
904extern void ceph_cleanup_empty_realms(struct ceph_mds_client *mdsc);
905
906extern struct ceph_snapid_map *ceph_get_snapid_map(struct ceph_mds_client *mdsc,
907 u64 snap);
908extern void ceph_put_snapid_map(struct ceph_mds_client* mdsc,
909 struct ceph_snapid_map *sm);
910extern void ceph_trim_snapid_map(struct ceph_mds_client *mdsc);
911extern void ceph_cleanup_snapid_map(struct ceph_mds_client *mdsc);
912
913
914/*
915 * a cap_snap is "pending" if it is still awaiting an in-progress
916 * sync write (that may/may not still update size, mtime, etc.).
917 */
918static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci)
919{
920 return !list_empty(&ci->i_cap_snaps) &&
921 list_last_entry(&ci->i_cap_snaps, struct ceph_cap_snap,
922 ci_item)->writing;
923}
924
925/* inode.c */
926struct ceph_mds_reply_info_in;
927struct ceph_mds_reply_dirfrag;
928
929extern const struct inode_operations ceph_file_iops;
930
931extern struct inode *ceph_alloc_inode(struct super_block *sb);
932extern void ceph_evict_inode(struct inode *inode);
933extern void ceph_free_inode(struct inode *inode);
934
935extern struct inode *ceph_get_inode(struct super_block *sb,
936 struct ceph_vino vino);
937extern struct inode *ceph_get_snapdir(struct inode *parent);
938extern int ceph_fill_file_size(struct inode *inode, int issued,
939 u32 truncate_seq, u64 truncate_size, u64 size);
940extern void ceph_fill_file_time(struct inode *inode, int issued,
941 u64 time_warp_seq, struct timespec64 *ctime,
942 struct timespec64 *mtime,
943 struct timespec64 *atime);
944extern int ceph_fill_inode(struct inode *inode, struct page *locked_page,
945 struct ceph_mds_reply_info_in *iinfo,
946 struct ceph_mds_reply_dirfrag *dirinfo,
947 struct ceph_mds_session *session, int cap_fmode,
948 struct ceph_cap_reservation *caps_reservation);
949extern int ceph_fill_trace(struct super_block *sb,
950 struct ceph_mds_request *req);
951extern int ceph_readdir_prepopulate(struct ceph_mds_request *req,
952 struct ceph_mds_session *session);
953
954extern int ceph_inode_holds_cap(struct inode *inode, int mask);
955
956extern bool ceph_inode_set_size(struct inode *inode, loff_t size);
957extern void __ceph_do_pending_vmtruncate(struct inode *inode);
958extern void ceph_queue_vmtruncate(struct inode *inode);
959extern void ceph_queue_invalidate(struct inode *inode);
960extern void ceph_queue_writeback(struct inode *inode);
961extern void ceph_async_iput(struct inode *inode);
962
963extern int __ceph_do_getattr(struct inode *inode, struct page *locked_page,
964 int mask, bool force);
965static inline int ceph_do_getattr(struct inode *inode, int mask, bool force)
966{
967 return __ceph_do_getattr(inode, NULL, mask, force);
968}
969extern int ceph_permission(struct inode *inode, int mask);
970extern int __ceph_setattr(struct inode *inode, struct iattr *attr);
971extern int ceph_setattr(struct dentry *dentry, struct iattr *attr);
972extern int ceph_getattr(const struct path *path, struct kstat *stat,
973 u32 request_mask, unsigned int flags);
974
975/* xattr.c */
976int __ceph_setxattr(struct inode *, const char *, const void *, size_t, int);
977ssize_t __ceph_getxattr(struct inode *, const char *, void *, size_t);
978extern ssize_t ceph_listxattr(struct dentry *, char *, size_t);
979extern struct ceph_buffer *__ceph_build_xattrs_blob(struct ceph_inode_info *ci);
980extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci);
981extern const struct xattr_handler *ceph_xattr_handlers[];
982
983struct ceph_acl_sec_ctx {
984#ifdef CONFIG_CEPH_FS_POSIX_ACL
985 void *default_acl;
986 void *acl;
987#endif
988#ifdef CONFIG_CEPH_FS_SECURITY_LABEL
989 void *sec_ctx;
990 u32 sec_ctxlen;
991#endif
992 struct ceph_pagelist *pagelist;
993};
994
995#ifdef CONFIG_SECURITY
996extern bool ceph_security_xattr_deadlock(struct inode *in);
997extern bool ceph_security_xattr_wanted(struct inode *in);
998#else
999static inline bool ceph_security_xattr_deadlock(struct inode *in)
1000{
1001 return false;
1002}
1003static inline bool ceph_security_xattr_wanted(struct inode *in)
1004{
1005 return false;
1006}
1007#endif
1008
1009#ifdef CONFIG_CEPH_FS_SECURITY_LABEL
1010extern int ceph_security_init_secctx(struct dentry *dentry, umode_t mode,
1011 struct ceph_acl_sec_ctx *ctx);
1012static inline void ceph_security_invalidate_secctx(struct inode *inode)
1013{
1014 security_inode_invalidate_secctx(inode);
1015}
1016#else
1017static inline int ceph_security_init_secctx(struct dentry *dentry, umode_t mode,
1018 struct ceph_acl_sec_ctx *ctx)
1019{
1020 return 0;
1021}
1022static inline void ceph_security_invalidate_secctx(struct inode *inode)
1023{
1024}
1025#endif
1026
1027void ceph_release_acl_sec_ctx(struct ceph_acl_sec_ctx *as_ctx);
1028
1029/* acl.c */
1030#ifdef CONFIG_CEPH_FS_POSIX_ACL
1031
1032struct posix_acl *ceph_get_acl(struct inode *, int);
1033int ceph_set_acl(struct inode *inode, struct posix_acl *acl, int type);
1034int ceph_pre_init_acls(struct inode *dir, umode_t *mode,
1035 struct ceph_acl_sec_ctx *as_ctx);
1036void ceph_init_inode_acls(struct inode *inode,
1037 struct ceph_acl_sec_ctx *as_ctx);
1038
1039static inline void ceph_forget_all_cached_acls(struct inode *inode)
1040{
1041 forget_all_cached_acls(inode);
1042}
1043
1044#else
1045
1046#define ceph_get_acl NULL
1047#define ceph_set_acl NULL
1048
1049static inline int ceph_pre_init_acls(struct inode *dir, umode_t *mode,
1050 struct ceph_acl_sec_ctx *as_ctx)
1051{
1052 return 0;
1053}
1054static inline void ceph_init_inode_acls(struct inode *inode,
1055 struct ceph_acl_sec_ctx *as_ctx)
1056{
1057}
1058static inline int ceph_acl_chmod(struct dentry *dentry, struct inode *inode)
1059{
1060 return 0;
1061}
1062
1063static inline void ceph_forget_all_cached_acls(struct inode *inode)
1064{
1065}
1066
1067#endif
1068
1069/* caps.c */
1070extern const char *ceph_cap_string(int c);
1071extern void ceph_handle_caps(struct ceph_mds_session *session,
1072 struct ceph_msg *msg);
1073extern struct ceph_cap *ceph_get_cap(struct ceph_mds_client *mdsc,
1074 struct ceph_cap_reservation *ctx);
1075extern void ceph_add_cap(struct inode *inode,
1076 struct ceph_mds_session *session, u64 cap_id,
1077 unsigned issued, unsigned wanted,
1078 unsigned cap, unsigned seq, u64 realmino, int flags,
1079 struct ceph_cap **new_cap);
1080extern void __ceph_remove_cap(struct ceph_cap *cap, bool queue_release);
1081extern void __ceph_remove_caps(struct ceph_inode_info *ci);
1082extern void ceph_put_cap(struct ceph_mds_client *mdsc,
1083 struct ceph_cap *cap);
1084extern int ceph_is_any_caps(struct inode *inode);
1085
1086extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc);
1087extern int ceph_fsync(struct file *file, loff_t start, loff_t end,
1088 int datasync);
1089extern void ceph_early_kick_flushing_caps(struct ceph_mds_client *mdsc,
1090 struct ceph_mds_session *session);
1091extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc,
1092 struct ceph_mds_session *session);
1093void ceph_kick_flushing_inode_caps(struct ceph_mds_session *session,
1094 struct ceph_inode_info *ci);
1095extern struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci,
1096 int mds);
1097extern void ceph_take_cap_refs(struct ceph_inode_info *ci, int caps,
1098 bool snap_rwsem_locked);
1099extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps);
1100extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had);
1101extern void ceph_put_cap_refs_no_check_caps(struct ceph_inode_info *ci,
1102 int had);
1103extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
1104 struct ceph_snap_context *snapc);
1105extern void ceph_flush_snaps(struct ceph_inode_info *ci,
1106 struct ceph_mds_session **psession);
1107extern bool __ceph_should_report_size(struct ceph_inode_info *ci);
1108extern void ceph_check_caps(struct ceph_inode_info *ci, int flags,
1109 struct ceph_mds_session *session);
1110extern void ceph_check_delayed_caps(struct ceph_mds_client *mdsc);
1111extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc);
1112extern int ceph_drop_caps_for_unlink(struct inode *inode);
1113extern int ceph_encode_inode_release(void **p, struct inode *inode,
1114 int mds, int drop, int unless, int force);
1115extern int ceph_encode_dentry_release(void **p, struct dentry *dn,
1116 struct inode *dir,
1117 int mds, int drop, int unless);
1118
1119extern int ceph_get_caps(struct file *filp, int need, int want,
1120 loff_t endoff, int *got, struct page **pinned_page);
1121extern int ceph_try_get_caps(struct inode *inode,
1122 int need, int want, bool nonblock, int *got);
1123
1124/* for counting open files by mode */
1125extern void ceph_get_fmode(struct ceph_inode_info *ci, int mode, int count);
1126extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode, int count);
1127extern void __ceph_touch_fmode(struct ceph_inode_info *ci,
1128 struct ceph_mds_client *mdsc, int fmode);
1129
1130/* addr.c */
1131extern const struct address_space_operations ceph_aops;
1132extern int ceph_mmap(struct file *file, struct vm_area_struct *vma);
1133extern int ceph_uninline_data(struct file *filp, struct page *locked_page);
1134extern int ceph_pool_perm_check(struct inode *inode, int need);
1135extern void ceph_pool_perm_destroy(struct ceph_mds_client* mdsc);
1136
1137/* file.c */
1138extern const struct file_operations ceph_file_fops;
1139
1140extern int ceph_renew_caps(struct inode *inode, int fmode);
1141extern int ceph_open(struct inode *inode, struct file *file);
1142extern int ceph_atomic_open(struct inode *dir, struct dentry *dentry,
1143 struct file *file, unsigned flags, umode_t mode);
1144extern int ceph_release(struct inode *inode, struct file *filp);
1145extern void ceph_fill_inline_data(struct inode *inode, struct page *locked_page,
1146 char *data, size_t len);
1147
1148/* dir.c */
1149extern const struct file_operations ceph_dir_fops;
1150extern const struct file_operations ceph_snapdir_fops;
1151extern const struct inode_operations ceph_dir_iops;
1152extern const struct inode_operations ceph_snapdir_iops;
1153extern const struct dentry_operations ceph_dentry_ops;
1154
1155extern loff_t ceph_make_fpos(unsigned high, unsigned off, bool hash_order);
1156extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry);
1157extern int ceph_handle_snapdir(struct ceph_mds_request *req,
1158 struct dentry *dentry, int err);
1159extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req,
1160 struct dentry *dentry, int err);
1161
1162extern void __ceph_dentry_lease_touch(struct ceph_dentry_info *di);
1163extern void __ceph_dentry_dir_lease_touch(struct ceph_dentry_info *di);
1164extern void ceph_invalidate_dentry_lease(struct dentry *dentry);
1165extern int ceph_trim_dentries(struct ceph_mds_client *mdsc);
1166extern unsigned ceph_dentry_hash(struct inode *dir, struct dentry *dn);
1167extern void ceph_readdir_cache_release(struct ceph_readdir_cache_control *ctl);
1168
1169/* ioctl.c */
1170extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
1171
1172/* export.c */
1173extern const struct export_operations ceph_export_ops;
1174struct inode *ceph_lookup_inode(struct super_block *sb, u64 ino);
1175
1176/* locks.c */
1177extern __init void ceph_flock_init(void);
1178extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl);
1179extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl);
1180extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num);
1181extern int ceph_encode_locks_to_buffer(struct inode *inode,
1182 struct ceph_filelock *flocks,
1183 int num_fcntl_locks,
1184 int num_flock_locks);
1185extern int ceph_locks_to_pagelist(struct ceph_filelock *flocks,
1186 struct ceph_pagelist *pagelist,
1187 int num_fcntl_locks, int num_flock_locks);
1188
1189/* debugfs.c */
1190extern void ceph_fs_debugfs_init(struct ceph_fs_client *client);
1191extern void ceph_fs_debugfs_cleanup(struct ceph_fs_client *client);
1192
1193/* quota.c */
1194static inline bool __ceph_has_any_quota(struct ceph_inode_info *ci)
1195{
1196 return ci->i_max_files || ci->i_max_bytes;
1197}
1198
1199extern void ceph_adjust_quota_realms_count(struct inode *inode, bool inc);
1200
1201static inline void __ceph_update_quota(struct ceph_inode_info *ci,
1202 u64 max_bytes, u64 max_files)
1203{
1204 bool had_quota, has_quota;
1205 had_quota = __ceph_has_any_quota(ci);
1206 ci->i_max_bytes = max_bytes;
1207 ci->i_max_files = max_files;
1208 has_quota = __ceph_has_any_quota(ci);
1209
1210 if (had_quota != has_quota)
1211 ceph_adjust_quota_realms_count(&ci->vfs_inode, has_quota);
1212}
1213
1214extern void ceph_handle_quota(struct ceph_mds_client *mdsc,
1215 struct ceph_mds_session *session,
1216 struct ceph_msg *msg);
1217extern bool ceph_quota_is_max_files_exceeded(struct inode *inode);
1218extern bool ceph_quota_is_max_bytes_exceeded(struct inode *inode,
1219 loff_t newlen);
1220extern bool ceph_quota_is_max_bytes_approaching(struct inode *inode,
1221 loff_t newlen);
1222extern bool ceph_quota_update_statfs(struct ceph_fs_client *fsc,
1223 struct kstatfs *buf);
1224extern int ceph_quota_check_rename(struct ceph_mds_client *mdsc,
1225 struct inode *old, struct inode *new);
1226extern void ceph_cleanup_quotarealms_inodes(struct ceph_mds_client *mdsc);
1227
1228#endif /* _FS_CEPH_SUPER_H */