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