blob: ff7aaa32736c28cedfa113efb3c7ed9620fb90d6 [file] [log] [blame]
Sage Weilde576062009-10-06 11:31:07 -07001#ifndef _FS_CEPH_SUPER_H
2#define _FS_CEPH_SUPER_H
3
4#include "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
15#include "types.h"
16#include "messenger.h"
17#include "msgpool.h"
18#include "mon_client.h"
19#include "mds_client.h"
20#include "osd_client.h"
21#include "ceph_fs.h"
22
23/* f_type in struct statfs */
24#define CEPH_SUPER_MAGIC 0x00c36400
25
26/* large granularity for statfs utilization stats to facilitate
27 * large volume sizes on 32-bit machines. */
28#define CEPH_BLOCK_SHIFT 20 /* 1 MB */
29#define CEPH_BLOCK (1 << CEPH_BLOCK_SHIFT)
30
31/*
32 * mount options
33 */
34#define CEPH_OPT_FSID (1<<0)
35#define CEPH_OPT_NOSHARE (1<<1) /* don't share client with other sbs */
36#define CEPH_OPT_MYIP (1<<2) /* specified my ip */
37#define CEPH_OPT_DIRSTAT (1<<4) /* funky `cat dirname` for stats */
38#define CEPH_OPT_RBYTES (1<<5) /* dir st_bytes = rbytes */
39#define CEPH_OPT_NOCRC (1<<6) /* no data crc on writes */
40#define CEPH_OPT_NOASYNCREADDIR (1<<7) /* no dcache readdir */
41
42#define CEPH_OPT_DEFAULT (CEPH_OPT_RBYTES)
43
44#define ceph_set_opt(client, opt) \
Sage Weil6b805182009-10-27 11:50:50 -070045 (client)->mount_args->flags |= CEPH_OPT_##opt;
Sage Weilde576062009-10-06 11:31:07 -070046#define ceph_test_opt(client, opt) \
Sage Weil6b805182009-10-27 11:50:50 -070047 (!!((client)->mount_args->flags & CEPH_OPT_##opt))
Sage Weilde576062009-10-06 11:31:07 -070048
49
Sage Weilde576062009-10-06 11:31:07 -070050struct ceph_mount_args {
51 int sb_flags;
Sage Weil6b805182009-10-27 11:50:50 -070052 int num_mon;
53 struct ceph_entity_addr *mon_addr;
Sage Weilde576062009-10-06 11:31:07 -070054 int flags;
55 int mount_timeout;
Yehuda Sadehf5a20412010-02-03 11:00:26 -080056 int osd_idle_ttl;
Sage Weilde576062009-10-06 11:31:07 -070057 int caps_wanted_delay_min, caps_wanted_delay_max;
58 struct ceph_fsid fsid;
59 struct ceph_entity_addr my_addr;
60 int wsize;
61 int rsize; /* max readahead */
62 int max_readdir; /* max readdir size */
Yehuda Sadeh2baba252009-12-18 13:51:57 -080063 int congestion_kb; /* max readdir size */
Sage Weilde576062009-10-06 11:31:07 -070064 int osd_timeout;
65 char *snapdir_name; /* default ".snap" */
Sage Weil4e7a5dc2009-11-18 16:19:57 -080066 char *name;
Sage Weilde576062009-10-06 11:31:07 -070067 char *secret;
68 int cap_release_safety;
69};
70
71/*
72 * defaults
73 */
74#define CEPH_MOUNT_TIMEOUT_DEFAULT 60
Yehuda Sadehf5a20412010-02-03 11:00:26 -080075#define CEPH_OSD_IDLE_TTL_DEFAULT 60
Sage Weil8fa97652009-10-16 14:44:35 -070076#define CEPH_MOUNT_RSIZE_DEFAULT (512*1024) /* readahead */
Sage Weilde576062009-10-06 11:31:07 -070077
78#define CEPH_MSG_MAX_FRONT_LEN (16*1024*1024)
79#define CEPH_MSG_MAX_DATA_LEN (16*1024*1024)
80
81#define CEPH_SNAPDIRNAME_DEFAULT ".snap"
Sage Weil4e7a5dc2009-11-18 16:19:57 -080082#define CEPH_AUTH_NAME_DEFAULT "guest"
Sage Weilde576062009-10-06 11:31:07 -070083
84/*
85 * Delay telling the MDS we no longer want caps, in case we reopen
86 * the file. Delay a minimum amount of time, even if we send a cap
87 * message for some other reason. Otherwise, take the oppotunity to
88 * update the mds to avoid sending another message later.
89 */
90#define CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT 5 /* cap release delay */
91#define CEPH_CAPS_WANTED_DELAY_MAX_DEFAULT 60 /* cap release delay */
92
93
94/* mount state */
95enum {
96 CEPH_MOUNT_MOUNTING,
97 CEPH_MOUNT_MOUNTED,
98 CEPH_MOUNT_UNMOUNTING,
99 CEPH_MOUNT_UNMOUNTED,
100 CEPH_MOUNT_SHUTDOWN,
101};
102
103/*
104 * subtract jiffies
105 */
106static inline unsigned long time_sub(unsigned long a, unsigned long b)
107{
108 BUG_ON(time_after(b, a));
109 return (long)a - (long)b;
110}
111
112/*
113 * per-filesystem client state
114 *
115 * possibly shared by multiple mount points, if they are
116 * mounting the same ceph filesystem/cluster.
117 */
118struct ceph_client {
Sage Weil07433042009-11-18 16:50:41 -0800119 struct ceph_fsid fsid;
120 bool have_fsid;
Sage Weilde576062009-10-06 11:31:07 -0700121
122 struct mutex mount_mutex; /* serialize mount attempts */
Sage Weil6b805182009-10-27 11:50:50 -0700123 struct ceph_mount_args *mount_args;
Sage Weilde576062009-10-06 11:31:07 -0700124
125 struct super_block *sb;
126
127 unsigned long mount_state;
Sage Weil9bd2e6f2010-02-02 16:21:06 -0800128 wait_queue_head_t auth_wq;
Sage Weilde576062009-10-06 11:31:07 -0700129
Sage Weil9bd2e6f2010-02-02 16:21:06 -0800130 int auth_err;
Sage Weilde576062009-10-06 11:31:07 -0700131
Sage Weil85ccce42010-02-17 10:02:43 -0800132 int min_caps; /* min caps i added */
133
Sage Weilde576062009-10-06 11:31:07 -0700134 struct ceph_messenger *msgr; /* messenger instance */
135 struct ceph_mon_client monc;
136 struct ceph_mds_client mdsc;
137 struct ceph_osd_client osdc;
138
139 /* writeback */
140 mempool_t *wb_pagevec_pool;
141 struct workqueue_struct *wb_wq;
142 struct workqueue_struct *pg_inv_wq;
143 struct workqueue_struct *trunc_wq;
Yehuda Sadeh2baba252009-12-18 13:51:57 -0800144 atomic_long_t writeback_count;
Sage Weilde576062009-10-06 11:31:07 -0700145
146 struct backing_dev_info backing_dev_info;
Sage Weil07433042009-11-18 16:50:41 -0800147
148#ifdef CONFIG_DEBUG_FS
149 struct dentry *debugfs_monmap;
150 struct dentry *debugfs_mdsmap, *debugfs_osdmap;
151 struct dentry *debugfs_dir, *debugfs_dentry_lru, *debugfs_caps;
Yehuda Sadeh2baba252009-12-18 13:51:57 -0800152 struct dentry *debugfs_congestion_kb;
Sage Weil06edf042009-12-15 14:44:32 -0800153 struct dentry *debugfs_bdi;
Sage Weil07433042009-11-18 16:50:41 -0800154#endif
Sage Weilde576062009-10-06 11:31:07 -0700155};
156
157static inline struct ceph_client *ceph_client(struct super_block *sb)
158{
159 return sb->s_fs_info;
160}
161
162
163/*
164 * File i/o capability. This tracks shared state with the metadata
165 * server that allows us to cache or writeback attributes or to read
166 * and write data. For any given inode, we should have one or more
167 * capabilities, one issued by each metadata server, and our
168 * cumulative access is the OR of all issued capabilities.
169 *
170 * Each cap is referenced by the inode's i_caps rbtree and by per-mds
171 * session capability lists.
172 */
173struct ceph_cap {
174 struct ceph_inode_info *ci;
175 struct rb_node ci_node; /* per-ci cap tree */
176 struct ceph_mds_session *session;
177 struct list_head session_caps; /* per-session caplist */
178 int mds;
179 u64 cap_id; /* unique cap id (mds provided) */
180 int issued; /* latest, from the mds */
181 int implemented; /* implemented superset of issued (for revocation) */
182 int mds_wanted;
Sage Weil685f9a5d2009-11-09 12:05:48 -0800183 u32 seq, issue_seq, mseq;
184 u32 cap_gen; /* active/stale cycle */
Sage Weilde576062009-10-06 11:31:07 -0700185 unsigned long last_used;
186 struct list_head caps_item;
187};
188
189#define CHECK_CAPS_NODELAY 1 /* do not delay any further */
190#define CHECK_CAPS_AUTHONLY 2 /* only check auth cap */
191#define CHECK_CAPS_FLUSH 4 /* flush any dirty caps */
192
193/*
194 * Snapped cap state that is pending flush to mds. When a snapshot occurs,
195 * we first complete any in-process sync writes and writeback any dirty
196 * data before flushing the snapped state (tracked here) back to the MDS.
197 */
198struct ceph_cap_snap {
199 atomic_t nref;
200 struct ceph_inode_info *ci;
201 struct list_head ci_item, flushing_item;
202
203 u64 follows, flush_tid;
204 int issued, dirty;
205 struct ceph_snap_context *context;
206
207 mode_t mode;
208 uid_t uid;
209 gid_t gid;
210
211 void *xattr_blob;
212 int xattr_len;
213 u64 xattr_version;
214
215 u64 size;
216 struct timespec mtime, atime, ctime;
217 u64 time_warp_seq;
218 int writing; /* a sync write is still in progress */
219 int dirty_pages; /* dirty pages awaiting writeback */
220};
221
222static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap)
223{
224 if (atomic_dec_and_test(&capsnap->nref))
225 kfree(capsnap);
226}
227
228/*
229 * The frag tree describes how a directory is fragmented, potentially across
230 * multiple metadata servers. It is also used to indicate points where
231 * metadata authority is delegated, and whether/where metadata is replicated.
232 *
233 * A _leaf_ frag will be present in the i_fragtree IFF there is
234 * delegation info. That is, if mds >= 0 || ndist > 0.
235 */
236#define CEPH_MAX_DIRFRAG_REP 4
237
238struct ceph_inode_frag {
239 struct rb_node node;
240
241 /* fragtree state */
242 u32 frag;
243 int split_by; /* i.e. 2^(split_by) children */
244
245 /* delegation and replication info */
246 int mds; /* -1 if same authority as parent */
247 int ndist; /* >0 if replicated */
248 int dist[CEPH_MAX_DIRFRAG_REP];
249};
250
251/*
252 * We cache inode xattrs as an encoded blob until they are first used,
253 * at which point we parse them into an rbtree.
254 */
255struct ceph_inode_xattr {
256 struct rb_node node;
257
258 const char *name;
259 int name_len;
260 const char *val;
261 int val_len;
262 int dirty;
263
264 int should_free_name;
265 int should_free_val;
266};
267
268struct ceph_inode_xattrs_info {
269 /*
270 * (still encoded) xattr blob. we avoid the overhead of parsing
271 * this until someone actually calls getxattr, etc.
272 *
273 * blob->vec.iov_len == 4 implies there are no xattrs; blob ==
274 * NULL means we don't know.
275 */
276 struct ceph_buffer *blob, *prealloc_blob;
277
278 struct rb_root index;
279 bool dirty;
280 int count;
281 int names_size;
282 int vals_size;
283 u64 version, index_version;
284};
285
286/*
287 * Ceph inode.
288 */
289#define CEPH_I_COMPLETE 1 /* we have complete directory cached */
290#define CEPH_I_NODELAY 4 /* do not delay cap release */
291#define CEPH_I_FLUSH 8 /* do not delay flush of dirty metadata */
292
293struct ceph_inode_info {
294 struct ceph_vino i_vino; /* ceph ino + snap */
295
296 u64 i_version;
297 u32 i_time_warp_seq;
298
299 unsigned i_ceph_flags;
300 unsigned long i_release_count;
301
302 struct ceph_file_layout i_layout;
303 char *i_symlink;
304
305 /* for dirs */
306 struct timespec i_rctime;
307 u64 i_rbytes, i_rfiles, i_rsubdirs;
308 u64 i_files, i_subdirs;
309 u64 i_max_offset; /* largest readdir offset, set with I_COMPLETE */
310
311 struct rb_root i_fragtree;
312 struct mutex i_fragtree_mutex;
313
314 struct ceph_inode_xattrs_info i_xattrs;
315
316 /* capabilities. protected _both_ by i_lock and cap->session's
317 * s_mutex. */
318 struct rb_root i_caps; /* cap list */
319 struct ceph_cap *i_auth_cap; /* authoritative cap, if any */
320 unsigned i_dirty_caps, i_flushing_caps; /* mask of dirtied fields */
321 struct list_head i_dirty_item, i_flushing_item;
322 u64 i_cap_flush_seq;
323 /* we need to track cap writeback on a per-cap-bit basis, to allow
324 * overlapping, pipelined cap flushes to the mds. we can probably
325 * reduce the tid to 8 bits if we're concerned about inode size. */
326 u16 i_cap_flush_last_tid, i_cap_flush_tid[CEPH_CAP_BITS];
327 wait_queue_head_t i_cap_wq; /* threads waiting on a capability */
328 unsigned long i_hold_caps_min; /* jiffies */
329 unsigned long i_hold_caps_max; /* jiffies */
330 struct list_head i_cap_delay_list; /* for delayed cap release to mds */
331 int i_cap_exporting_mds; /* to handle cap migration between */
332 unsigned i_cap_exporting_mseq; /* mds's. */
333 unsigned i_cap_exporting_issued;
334 struct ceph_cap_reservation i_cap_migration_resv;
335 struct list_head i_cap_snaps; /* snapped state pending flush to mds */
336 struct ceph_snap_context *i_head_snapc; /* set if wr_buffer_head > 0 */
337 unsigned i_snap_caps; /* cap bits for snapped files */
338
339 int i_nr_by_mode[CEPH_FILE_MODE_NUM]; /* open file counts */
340
341 u32 i_truncate_seq; /* last truncate to smaller size */
342 u64 i_truncate_size; /* and the size we last truncated down to */
343 int i_truncate_pending; /* still need to call vmtruncate */
344
345 u64 i_max_size; /* max file size authorized by mds */
346 u64 i_reported_size; /* (max_)size reported to or requested of mds */
347 u64 i_wanted_max_size; /* offset we'd like to write too */
348 u64 i_requested_max_size; /* max_size we've requested */
349
350 /* held references to caps */
351 int i_pin_ref;
352 int i_rd_ref, i_rdcache_ref, i_wr_ref;
353 int i_wrbuffer_ref, i_wrbuffer_ref_head;
354 u32 i_shared_gen; /* increment each time we get FILE_SHARED */
355 u32 i_rdcache_gen; /* we increment this each time we get
356 FILE_CACHE. If it's non-zero, we
357 _may_ have cached pages. */
358 u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */
359
360 struct list_head i_unsafe_writes; /* uncommitted sync writes */
361 struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */
362 spinlock_t i_unsafe_lock;
363
364 struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */
365 int i_snap_realm_counter; /* snap realm (if caps) */
366 struct list_head i_snap_realm_item;
367 struct list_head i_snap_flush_item;
368
369 struct work_struct i_wb_work; /* writeback work */
370 struct work_struct i_pg_inv_work; /* page invalidation work */
371
372 struct work_struct i_vmtruncate_work;
373
374 struct inode vfs_inode; /* at end */
375};
376
377static inline struct ceph_inode_info *ceph_inode(struct inode *inode)
378{
Noah Watkinsfbbccec2009-10-28 11:54:49 -0700379 return container_of(inode, struct ceph_inode_info, vfs_inode);
Sage Weilde576062009-10-06 11:31:07 -0700380}
381
382static inline void ceph_i_clear(struct inode *inode, unsigned mask)
383{
384 struct ceph_inode_info *ci = ceph_inode(inode);
385
386 spin_lock(&inode->i_lock);
387 ci->i_ceph_flags &= ~mask;
388 spin_unlock(&inode->i_lock);
389}
390
391static inline void ceph_i_set(struct inode *inode, unsigned mask)
392{
393 struct ceph_inode_info *ci = ceph_inode(inode);
394
395 spin_lock(&inode->i_lock);
396 ci->i_ceph_flags |= mask;
397 spin_unlock(&inode->i_lock);
398}
399
400static inline bool ceph_i_test(struct inode *inode, unsigned mask)
401{
402 struct ceph_inode_info *ci = ceph_inode(inode);
403 bool r;
404
405 smp_mb();
406 r = (ci->i_ceph_flags & mask) == mask;
407 return r;
408}
409
410
411/* find a specific frag @f */
412extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci,
413 u32 f);
414
415/*
416 * choose fragment for value @v. copy frag content to pfrag, if leaf
417 * exists
418 */
419extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v,
420 struct ceph_inode_frag *pfrag,
421 int *found);
422
423/*
424 * Ceph dentry state
425 */
426struct ceph_dentry_info {
427 struct ceph_mds_session *lease_session;
428 u32 lease_gen, lease_shared_gen;
429 u32 lease_seq;
430 unsigned long lease_renew_after, lease_renew_from;
431 struct list_head lru;
432 struct dentry *dentry;
433 u64 time;
434 u64 offset;
435};
436
437static inline struct ceph_dentry_info *ceph_dentry(struct dentry *dentry)
438{
439 return (struct ceph_dentry_info *)dentry->d_fsdata;
440}
441
442static inline loff_t ceph_make_fpos(unsigned frag, unsigned off)
443{
444 return ((loff_t)frag << 32) | (loff_t)off;
445}
446
447/*
448 * ino_t is <64 bits on many architectures, blech.
449 *
450 * don't include snap in ino hash, at least for now.
451 */
452static inline ino_t ceph_vino_to_ino(struct ceph_vino vino)
453{
454 ino_t ino = (ino_t)vino.ino; /* ^ (vino.snap << 20); */
455#if BITS_PER_LONG == 32
456 ino ^= vino.ino >> (sizeof(u64)-sizeof(ino_t)) * 8;
457 if (!ino)
458 ino = 1;
459#endif
460 return ino;
461}
462
463static inline int ceph_set_ino_cb(struct inode *inode, void *data)
464{
465 ceph_inode(inode)->i_vino = *(struct ceph_vino *)data;
466 inode->i_ino = ceph_vino_to_ino(*(struct ceph_vino *)data);
467 return 0;
468}
469
470static inline struct ceph_vino ceph_vino(struct inode *inode)
471{
472 return ceph_inode(inode)->i_vino;
473}
474
475/* for printf-style formatting */
476#define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap
477
478static inline u64 ceph_ino(struct inode *inode)
479{
480 return ceph_inode(inode)->i_vino.ino;
481}
482static inline u64 ceph_snap(struct inode *inode)
483{
484 return ceph_inode(inode)->i_vino.snap;
485}
486
487static inline int ceph_ino_compare(struct inode *inode, void *data)
488{
489 struct ceph_vino *pvino = (struct ceph_vino *)data;
490 struct ceph_inode_info *ci = ceph_inode(inode);
491 return ci->i_vino.ino == pvino->ino &&
492 ci->i_vino.snap == pvino->snap;
493}
494
495static inline struct inode *ceph_find_inode(struct super_block *sb,
496 struct ceph_vino vino)
497{
498 ino_t t = ceph_vino_to_ino(vino);
499 return ilookup5(sb, t, ceph_ino_compare, &vino);
500}
501
502
503/*
504 * caps helpers
505 */
506static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci)
507{
508 return !RB_EMPTY_ROOT(&ci->i_caps);
509}
510
511extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented);
512extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t);
513extern int __ceph_caps_issued_other(struct ceph_inode_info *ci,
514 struct ceph_cap *cap);
515
516static inline int ceph_caps_issued(struct ceph_inode_info *ci)
517{
518 int issued;
519 spin_lock(&ci->vfs_inode.i_lock);
520 issued = __ceph_caps_issued(ci, NULL);
521 spin_unlock(&ci->vfs_inode.i_lock);
522 return issued;
523}
524
525static inline int ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask,
526 int touch)
527{
528 int r;
529 spin_lock(&ci->vfs_inode.i_lock);
530 r = __ceph_caps_issued_mask(ci, mask, touch);
531 spin_unlock(&ci->vfs_inode.i_lock);
532 return r;
533}
534
535static inline int __ceph_caps_dirty(struct ceph_inode_info *ci)
536{
537 return ci->i_dirty_caps | ci->i_flushing_caps;
538}
Sage Weilafcdaea2009-10-14 14:27:38 -0700539extern void __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask);
Sage Weilde576062009-10-06 11:31:07 -0700540
541extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask);
542extern int __ceph_caps_used(struct ceph_inode_info *ci);
543
544extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci);
545
546/*
547 * wanted, by virtue of open file modes AND cap refs (buffered/cached data)
548 */
549static inline int __ceph_caps_wanted(struct ceph_inode_info *ci)
550{
551 int w = __ceph_caps_file_wanted(ci) | __ceph_caps_used(ci);
552 if (w & CEPH_CAP_FILE_BUFFER)
553 w |= CEPH_CAP_FILE_EXCL; /* we want EXCL if dirty data */
554 return w;
555}
556
557/* what the mds thinks we want */
558extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci);
559
560extern void ceph_caps_init(void);
561extern void ceph_caps_finalize(void);
Sage Weil85ccce42010-02-17 10:02:43 -0800562extern void ceph_adjust_min_caps(int delta);
Sage Weilde576062009-10-06 11:31:07 -0700563extern int ceph_reserve_caps(struct ceph_cap_reservation *ctx, int need);
564extern int ceph_unreserve_caps(struct ceph_cap_reservation *ctx);
565extern void ceph_reservation_status(struct ceph_client *client,
566 int *total, int *avail, int *used,
Sage Weil85ccce42010-02-17 10:02:43 -0800567 int *reserved, int *min);
Sage Weilde576062009-10-06 11:31:07 -0700568
569static inline struct ceph_client *ceph_inode_to_client(struct inode *inode)
570{
571 return (struct ceph_client *)inode->i_sb->s_fs_info;
572}
573
574static inline struct ceph_client *ceph_sb_to_client(struct super_block *sb)
575{
576 return (struct ceph_client *)sb->s_fs_info;
577}
578
Sage Weilde576062009-10-06 11:31:07 -0700579
580/*
581 * we keep buffered readdir results attached to file->private_data
582 */
583struct ceph_file_info {
584 int fmode; /* initialized on open */
585
586 /* readdir: position within the dir */
587 u32 frag;
588 struct ceph_mds_request *last_readdir;
589 int at_end;
590
591 /* readdir: position within a frag */
592 unsigned offset; /* offset of last chunk, adjusted for . and .. */
593 u64 next_offset; /* offset of next chunk (last_name's + 1) */
594 char *last_name; /* last entry in previous chunk */
595 struct dentry *dentry; /* next dentry (for dcache readdir) */
596 unsigned long dir_release_count;
597
598 /* used for -o dirstat read() on directory thing */
599 char *dir_info;
600 int dir_info_len;
601};
602
603
604
605/*
606 * snapshots
607 */
608
609/*
610 * A "snap context" is the set of existing snapshots when we
611 * write data. It is used by the OSD to guide its COW behavior.
612 *
613 * The ceph_snap_context is refcounted, and attached to each dirty
614 * page, indicating which context the dirty data belonged when it was
615 * dirtied.
616 */
617struct ceph_snap_context {
618 atomic_t nref;
619 u64 seq;
620 int num_snaps;
621 u64 snaps[];
622};
623
624static inline struct ceph_snap_context *
625ceph_get_snap_context(struct ceph_snap_context *sc)
626{
627 /*
628 printk("get_snap_context %p %d -> %d\n", sc, atomic_read(&sc->nref),
629 atomic_read(&sc->nref)+1);
630 */
631 if (sc)
632 atomic_inc(&sc->nref);
633 return sc;
634}
635
636static inline void ceph_put_snap_context(struct ceph_snap_context *sc)
637{
638 if (!sc)
639 return;
640 /*
641 printk("put_snap_context %p %d -> %d\n", sc, atomic_read(&sc->nref),
642 atomic_read(&sc->nref)-1);
643 */
644 if (atomic_dec_and_test(&sc->nref)) {
645 /*printk(" deleting snap_context %p\n", sc);*/
646 kfree(sc);
647 }
648}
649
650/*
651 * A "snap realm" describes a subset of the file hierarchy sharing
652 * the same set of snapshots that apply to it. The realms themselves
653 * are organized into a hierarchy, such that children inherit (some of)
654 * the snapshots of their parents.
655 *
656 * All inodes within the realm that have capabilities are linked into a
657 * per-realm list.
658 */
659struct ceph_snap_realm {
660 u64 ino;
661 atomic_t nref;
Sage Weila105f002010-02-15 14:37:55 -0800662 struct rb_node node;
663
Sage Weilde576062009-10-06 11:31:07 -0700664 u64 created, seq;
665 u64 parent_ino;
666 u64 parent_since; /* snapid when our current parent became so */
667
668 u64 *prior_parent_snaps; /* snaps inherited from any parents we */
669 int num_prior_parent_snaps; /* had prior to parent_since */
670 u64 *snaps; /* snaps specific to this realm */
671 int num_snaps;
672
673 struct ceph_snap_realm *parent;
674 struct list_head children; /* list of child realms */
675 struct list_head child_item;
676
677 struct list_head empty_item; /* if i have ref==0 */
678
679 /* the current set of snaps for this realm */
680 struct ceph_snap_context *cached_context;
681
682 struct list_head inodes_with_caps;
683 spinlock_t inodes_with_caps_lock;
684};
685
686
687
688/*
689 * calculate the number of pages a given length and offset map onto,
690 * if we align the data.
691 */
692static inline int calc_pages_for(u64 off, u64 len)
693{
694 return ((off+len+PAGE_CACHE_SIZE-1) >> PAGE_CACHE_SHIFT) -
695 (off >> PAGE_CACHE_SHIFT);
696}
697
698
699
700/* snap.c */
701struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc,
702 u64 ino);
703extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc,
704 struct ceph_snap_realm *realm);
705extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc,
706 struct ceph_snap_realm *realm);
707extern int ceph_update_snap_trace(struct ceph_mds_client *m,
708 void *p, void *e, bool deletion);
709extern void ceph_handle_snap(struct ceph_mds_client *mdsc,
Sage Weil2600d2d2010-02-22 15:12:16 -0800710 struct ceph_mds_session *session,
Sage Weilde576062009-10-06 11:31:07 -0700711 struct ceph_msg *msg);
712extern void ceph_queue_cap_snap(struct ceph_inode_info *ci,
713 struct ceph_snap_context *snapc);
714extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci,
715 struct ceph_cap_snap *capsnap);
716extern void ceph_cleanup_empty_realms(struct ceph_mds_client *mdsc);
717
718/*
719 * a cap_snap is "pending" if it is still awaiting an in-progress
720 * sync write (that may/may not still update size, mtime, etc.).
721 */
722static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci)
723{
724 return !list_empty(&ci->i_cap_snaps) &&
725 list_entry(ci->i_cap_snaps.prev, struct ceph_cap_snap,
726 ci_item)->writing;
727}
728
729
730/* super.c */
731extern struct kmem_cache *ceph_inode_cachep;
732extern struct kmem_cache *ceph_cap_cachep;
733extern struct kmem_cache *ceph_dentry_cachep;
734extern struct kmem_cache *ceph_file_cachep;
735
736extern const char *ceph_msg_type_name(int type);
Sage Weil07433042009-11-18 16:50:41 -0800737extern int ceph_check_fsid(struct ceph_client *client, struct ceph_fsid *fsid);
Sage Weilde576062009-10-06 11:31:07 -0700738
739#define FSID_FORMAT "%02x%02x%02x%02x-%02x%02x-%02x%02x-%02x%02x-" \
740 "%02x%02x%02x%02x%02x%02x"
741#define PR_FSID(f) (f)->fsid[0], (f)->fsid[1], (f)->fsid[2], (f)->fsid[3], \
742 (f)->fsid[4], (f)->fsid[5], (f)->fsid[6], (f)->fsid[7], \
743 (f)->fsid[8], (f)->fsid[9], (f)->fsid[10], (f)->fsid[11], \
744 (f)->fsid[12], (f)->fsid[13], (f)->fsid[14], (f)->fsid[15]
745
746/* inode.c */
747extern const struct inode_operations ceph_file_iops;
748
749extern struct inode *ceph_alloc_inode(struct super_block *sb);
750extern void ceph_destroy_inode(struct inode *inode);
751
752extern struct inode *ceph_get_inode(struct super_block *sb,
753 struct ceph_vino vino);
754extern struct inode *ceph_get_snapdir(struct inode *parent);
755extern int ceph_fill_file_size(struct inode *inode, int issued,
756 u32 truncate_seq, u64 truncate_size, u64 size);
757extern void ceph_fill_file_time(struct inode *inode, int issued,
758 u64 time_warp_seq, struct timespec *ctime,
759 struct timespec *mtime, struct timespec *atime);
760extern int ceph_fill_trace(struct super_block *sb,
761 struct ceph_mds_request *req,
762 struct ceph_mds_session *session);
763extern int ceph_readdir_prepopulate(struct ceph_mds_request *req,
764 struct ceph_mds_session *session);
765
766extern int ceph_inode_holds_cap(struct inode *inode, int mask);
767
768extern int ceph_inode_set_size(struct inode *inode, loff_t size);
Sage Weilde576062009-10-06 11:31:07 -0700769extern void __ceph_do_pending_vmtruncate(struct inode *inode);
Sage Weil3c6f6b72010-02-09 15:24:44 -0800770extern void ceph_queue_vmtruncate(struct inode *inode);
771
772extern void ceph_queue_invalidate(struct inode *inode);
773extern void ceph_queue_writeback(struct inode *inode);
Sage Weilde576062009-10-06 11:31:07 -0700774
775extern int ceph_do_getattr(struct inode *inode, int mask);
776extern int ceph_permission(struct inode *inode, int mask);
777extern int ceph_setattr(struct dentry *dentry, struct iattr *attr);
778extern int ceph_getattr(struct vfsmount *mnt, struct dentry *dentry,
779 struct kstat *stat);
780
781/* xattr.c */
782extern int ceph_setxattr(struct dentry *, const char *, const void *,
783 size_t, int);
784extern ssize_t ceph_getxattr(struct dentry *, const char *, void *, size_t);
785extern ssize_t ceph_listxattr(struct dentry *, char *, size_t);
786extern int ceph_removexattr(struct dentry *, const char *);
787extern void __ceph_build_xattrs_blob(struct ceph_inode_info *ci);
788extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci);
789
790/* caps.c */
791extern const char *ceph_cap_string(int c);
792extern void ceph_handle_caps(struct ceph_mds_session *session,
793 struct ceph_msg *msg);
794extern int ceph_add_cap(struct inode *inode,
795 struct ceph_mds_session *session, u64 cap_id,
796 int fmode, unsigned issued, unsigned wanted,
797 unsigned cap, unsigned seq, u64 realmino, int flags,
798 struct ceph_cap_reservation *caps_reservation);
Sage Weil7c1332b2010-02-16 11:39:45 -0800799extern void __ceph_remove_cap(struct ceph_cap *cap);
Sage Weilde576062009-10-06 11:31:07 -0700800static inline void ceph_remove_cap(struct ceph_cap *cap)
801{
802 struct inode *inode = &cap->ci->vfs_inode;
803 spin_lock(&inode->i_lock);
Sage Weil7c1332b2010-02-16 11:39:45 -0800804 __ceph_remove_cap(cap);
Sage Weilde576062009-10-06 11:31:07 -0700805 spin_unlock(&inode->i_lock);
806}
Sage Weil7c1332b2010-02-16 11:39:45 -0800807extern void ceph_put_cap(struct ceph_cap *cap);
Sage Weilde576062009-10-06 11:31:07 -0700808
809extern void ceph_queue_caps_release(struct inode *inode);
810extern int ceph_write_inode(struct inode *inode, int unused);
811extern int ceph_fsync(struct file *file, struct dentry *dentry, int datasync);
812extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc,
813 struct ceph_mds_session *session);
814extern int ceph_get_cap_mds(struct inode *inode);
815extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps);
816extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had);
817extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
818 struct ceph_snap_context *snapc);
819extern void __ceph_flush_snaps(struct ceph_inode_info *ci,
820 struct ceph_mds_session **psession);
821extern void ceph_check_caps(struct ceph_inode_info *ci, int flags,
822 struct ceph_mds_session *session);
Sage Weilafcdaea2009-10-14 14:27:38 -0700823extern void ceph_check_delayed_caps(struct ceph_mds_client *mdsc);
824extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc);
Sage Weilde576062009-10-06 11:31:07 -0700825
826extern int ceph_encode_inode_release(void **p, struct inode *inode,
827 int mds, int drop, int unless, int force);
828extern int ceph_encode_dentry_release(void **p, struct dentry *dn,
829 int mds, int drop, int unless);
830
831extern int ceph_get_caps(struct ceph_inode_info *ci, int need, int want,
832 int *got, loff_t endoff);
833
834/* for counting open files by mode */
835static inline void __ceph_get_fmode(struct ceph_inode_info *ci, int mode)
836{
837 ci->i_nr_by_mode[mode]++;
838}
839extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode);
840
841/* addr.c */
842extern const struct address_space_operations ceph_aops;
843extern int ceph_mmap(struct file *file, struct vm_area_struct *vma);
844
845/* file.c */
846extern const struct file_operations ceph_file_fops;
847extern const struct address_space_operations ceph_aops;
848extern int ceph_open(struct inode *inode, struct file *file);
849extern struct dentry *ceph_lookup_open(struct inode *dir, struct dentry *dentry,
850 struct nameidata *nd, int mode,
851 int locked_dir);
852extern int ceph_release(struct inode *inode, struct file *filp);
853extern void ceph_release_page_vector(struct page **pages, int num_pages);
854
855/* dir.c */
856extern const struct file_operations ceph_dir_fops;
857extern const struct inode_operations ceph_dir_iops;
858extern struct dentry_operations ceph_dentry_ops, ceph_snap_dentry_ops,
859 ceph_snapdir_dentry_ops;
860
861extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry);
862extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req,
863 struct dentry *dentry, int err);
864
865extern void ceph_dentry_lru_add(struct dentry *dn);
866extern void ceph_dentry_lru_touch(struct dentry *dn);
867extern void ceph_dentry_lru_del(struct dentry *dn);
868
869/*
870 * our d_ops vary depending on whether the inode is live,
871 * snapshotted (read-only), or a virtual ".snap" directory.
872 */
873int ceph_init_dentry(struct dentry *dentry);
874
875
876/* ioctl.c */
877extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
878
879/* export.c */
880extern const struct export_operations ceph_export_ops;
881
882/* debugfs.c */
883extern int ceph_debugfs_init(void);
884extern void ceph_debugfs_cleanup(void);
885extern int ceph_debugfs_client_init(struct ceph_client *client);
886extern void ceph_debugfs_client_cleanup(struct ceph_client *client);
887
888static inline struct inode *get_dentry_parent_inode(struct dentry *dentry)
889{
890 if (dentry && dentry->d_parent)
891 return dentry->d_parent->d_inode;
892
893 return NULL;
894}
895
896#endif /* _FS_CEPH_SUPER_H */