blob: 4c07acaf21e352b8e9a0b22300bcc18d4dc639c5 [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>
Stephen Rothwellf1a3d572010-01-18 11:53:08 +110014#include <linux/writeback.h>
Sage Weilde576062009-10-06 11:31:07 -070015
16#include "types.h"
17#include "messenger.h"
18#include "msgpool.h"
19#include "mon_client.h"
20#include "mds_client.h"
21#include "osd_client.h"
22#include "ceph_fs.h"
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 20 /* 1 MB */
30#define CEPH_BLOCK (1 << CEPH_BLOCK_SHIFT)
31
32/*
33 * mount options
34 */
35#define CEPH_OPT_FSID (1<<0)
36#define CEPH_OPT_NOSHARE (1<<1) /* don't share client with other sbs */
37#define CEPH_OPT_MYIP (1<<2) /* specified my ip */
38#define CEPH_OPT_DIRSTAT (1<<4) /* funky `cat dirname` for stats */
39#define CEPH_OPT_RBYTES (1<<5) /* dir st_bytes = rbytes */
40#define CEPH_OPT_NOCRC (1<<6) /* no data crc on writes */
41#define CEPH_OPT_NOASYNCREADDIR (1<<7) /* no dcache readdir */
42
43#define CEPH_OPT_DEFAULT (CEPH_OPT_RBYTES)
44
45#define ceph_set_opt(client, opt) \
Sage Weil6b805182009-10-27 11:50:50 -070046 (client)->mount_args->flags |= CEPH_OPT_##opt;
Sage Weilde576062009-10-06 11:31:07 -070047#define ceph_test_opt(client, opt) \
Sage Weil6b805182009-10-27 11:50:50 -070048 (!!((client)->mount_args->flags & CEPH_OPT_##opt))
Sage Weilde576062009-10-06 11:31:07 -070049
50
Sage Weilde576062009-10-06 11:31:07 -070051struct ceph_mount_args {
52 int sb_flags;
Sage Weil6b805182009-10-27 11:50:50 -070053 int num_mon;
54 struct ceph_entity_addr *mon_addr;
Sage Weilde576062009-10-06 11:31:07 -070055 int flags;
56 int mount_timeout;
Yehuda Sadehf5a20412010-02-03 11:00:26 -080057 int osd_idle_ttl;
Sage Weilde576062009-10-06 11:31:07 -070058 int caps_wanted_delay_min, caps_wanted_delay_max;
59 struct ceph_fsid fsid;
60 struct ceph_entity_addr my_addr;
61 int wsize;
62 int rsize; /* max readahead */
63 int max_readdir; /* max readdir size */
Yehuda Sadeh2baba252009-12-18 13:51:57 -080064 int congestion_kb; /* max readdir size */
Sage Weilde576062009-10-06 11:31:07 -070065 int osd_timeout;
Yehuda Sadeh422d2cb2010-02-26 15:32:31 -080066 int osd_keepalive_timeout;
Sage Weilde576062009-10-06 11:31:07 -070067 char *snapdir_name; /* default ".snap" */
Sage Weil4e7a5dc2009-11-18 16:19:57 -080068 char *name;
Sage Weilde576062009-10-06 11:31:07 -070069 char *secret;
70 int cap_release_safety;
71};
72
73/*
74 * defaults
75 */
76#define CEPH_MOUNT_TIMEOUT_DEFAULT 60
Yehuda Sadeh422d2cb2010-02-26 15:32:31 -080077#define CEPH_OSD_TIMEOUT_DEFAULT 60 /* seconds */
78#define CEPH_OSD_KEEPALIVE_DEFAULT 5
Yehuda Sadehf5a20412010-02-03 11:00:26 -080079#define CEPH_OSD_IDLE_TTL_DEFAULT 60
Sage Weil8fa97652009-10-16 14:44:35 -070080#define CEPH_MOUNT_RSIZE_DEFAULT (512*1024) /* readahead */
Sage Weilde576062009-10-06 11:31:07 -070081
82#define CEPH_MSG_MAX_FRONT_LEN (16*1024*1024)
83#define CEPH_MSG_MAX_DATA_LEN (16*1024*1024)
84
85#define CEPH_SNAPDIRNAME_DEFAULT ".snap"
Sage Weil4e7a5dc2009-11-18 16:19:57 -080086#define CEPH_AUTH_NAME_DEFAULT "guest"
Sage Weilde576062009-10-06 11:31:07 -070087
88/*
89 * Delay telling the MDS we no longer want caps, in case we reopen
90 * the file. Delay a minimum amount of time, even if we send a cap
91 * message for some other reason. Otherwise, take the oppotunity to
92 * update the mds to avoid sending another message later.
93 */
94#define CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT 5 /* cap release delay */
95#define CEPH_CAPS_WANTED_DELAY_MAX_DEFAULT 60 /* cap release delay */
96
97
98/* mount state */
99enum {
100 CEPH_MOUNT_MOUNTING,
101 CEPH_MOUNT_MOUNTED,
102 CEPH_MOUNT_UNMOUNTING,
103 CEPH_MOUNT_UNMOUNTED,
104 CEPH_MOUNT_SHUTDOWN,
105};
106
107/*
108 * subtract jiffies
109 */
110static inline unsigned long time_sub(unsigned long a, unsigned long b)
111{
112 BUG_ON(time_after(b, a));
113 return (long)a - (long)b;
114}
115
116/*
117 * per-filesystem client state
118 *
119 * possibly shared by multiple mount points, if they are
120 * mounting the same ceph filesystem/cluster.
121 */
122struct ceph_client {
Sage Weil07433042009-11-18 16:50:41 -0800123 struct ceph_fsid fsid;
124 bool have_fsid;
Sage Weilde576062009-10-06 11:31:07 -0700125
126 struct mutex mount_mutex; /* serialize mount attempts */
Sage Weil6b805182009-10-27 11:50:50 -0700127 struct ceph_mount_args *mount_args;
Sage Weilde576062009-10-06 11:31:07 -0700128
129 struct super_block *sb;
130
131 unsigned long mount_state;
Sage Weil9bd2e6f2010-02-02 16:21:06 -0800132 wait_queue_head_t auth_wq;
Sage Weilde576062009-10-06 11:31:07 -0700133
Sage Weil9bd2e6f2010-02-02 16:21:06 -0800134 int auth_err;
Sage Weilde576062009-10-06 11:31:07 -0700135
Sage Weil85ccce42010-02-17 10:02:43 -0800136 int min_caps; /* min caps i added */
137
Sage Weilde576062009-10-06 11:31:07 -0700138 struct ceph_messenger *msgr; /* messenger instance */
139 struct ceph_mon_client monc;
140 struct ceph_mds_client mdsc;
141 struct ceph_osd_client osdc;
142
143 /* writeback */
144 mempool_t *wb_pagevec_pool;
145 struct workqueue_struct *wb_wq;
146 struct workqueue_struct *pg_inv_wq;
147 struct workqueue_struct *trunc_wq;
Yehuda Sadeh2baba252009-12-18 13:51:57 -0800148 atomic_long_t writeback_count;
Sage Weilde576062009-10-06 11:31:07 -0700149
150 struct backing_dev_info backing_dev_info;
Sage Weil07433042009-11-18 16:50:41 -0800151
152#ifdef CONFIG_DEBUG_FS
153 struct dentry *debugfs_monmap;
154 struct dentry *debugfs_mdsmap, *debugfs_osdmap;
155 struct dentry *debugfs_dir, *debugfs_dentry_lru, *debugfs_caps;
Yehuda Sadeh2baba252009-12-18 13:51:57 -0800156 struct dentry *debugfs_congestion_kb;
Sage Weil06edf042009-12-15 14:44:32 -0800157 struct dentry *debugfs_bdi;
Sage Weil07433042009-11-18 16:50:41 -0800158#endif
Sage Weilde576062009-10-06 11:31:07 -0700159};
160
161static inline struct ceph_client *ceph_client(struct super_block *sb)
162{
163 return sb->s_fs_info;
164}
165
166
167/*
168 * File i/o capability. This tracks shared state with the metadata
169 * server that allows us to cache or writeback attributes or to read
170 * and write data. For any given inode, we should have one or more
171 * capabilities, one issued by each metadata server, and our
172 * cumulative access is the OR of all issued capabilities.
173 *
174 * Each cap is referenced by the inode's i_caps rbtree and by per-mds
175 * session capability lists.
176 */
177struct ceph_cap {
178 struct ceph_inode_info *ci;
179 struct rb_node ci_node; /* per-ci cap tree */
180 struct ceph_mds_session *session;
181 struct list_head session_caps; /* per-session caplist */
182 int mds;
183 u64 cap_id; /* unique cap id (mds provided) */
184 int issued; /* latest, from the mds */
185 int implemented; /* implemented superset of issued (for revocation) */
186 int mds_wanted;
Sage Weil685f9a5d2009-11-09 12:05:48 -0800187 u32 seq, issue_seq, mseq;
188 u32 cap_gen; /* active/stale cycle */
Sage Weilde576062009-10-06 11:31:07 -0700189 unsigned long last_used;
190 struct list_head caps_item;
191};
192
193#define CHECK_CAPS_NODELAY 1 /* do not delay any further */
194#define CHECK_CAPS_AUTHONLY 2 /* only check auth cap */
195#define CHECK_CAPS_FLUSH 4 /* flush any dirty caps */
196
197/*
198 * Snapped cap state that is pending flush to mds. When a snapshot occurs,
199 * we first complete any in-process sync writes and writeback any dirty
200 * data before flushing the snapped state (tracked here) back to the MDS.
201 */
202struct ceph_cap_snap {
203 atomic_t nref;
204 struct ceph_inode_info *ci;
205 struct list_head ci_item, flushing_item;
206
207 u64 follows, flush_tid;
208 int issued, dirty;
209 struct ceph_snap_context *context;
210
211 mode_t mode;
212 uid_t uid;
213 gid_t gid;
214
215 void *xattr_blob;
216 int xattr_len;
217 u64 xattr_version;
218
219 u64 size;
220 struct timespec mtime, atime, ctime;
221 u64 time_warp_seq;
222 int writing; /* a sync write is still in progress */
223 int dirty_pages; /* dirty pages awaiting writeback */
224};
225
226static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap)
227{
228 if (atomic_dec_and_test(&capsnap->nref))
229 kfree(capsnap);
230}
231
232/*
233 * The frag tree describes how a directory is fragmented, potentially across
234 * multiple metadata servers. It is also used to indicate points where
235 * metadata authority is delegated, and whether/where metadata is replicated.
236 *
237 * A _leaf_ frag will be present in the i_fragtree IFF there is
238 * delegation info. That is, if mds >= 0 || ndist > 0.
239 */
240#define CEPH_MAX_DIRFRAG_REP 4
241
242struct ceph_inode_frag {
243 struct rb_node node;
244
245 /* fragtree state */
246 u32 frag;
247 int split_by; /* i.e. 2^(split_by) children */
248
249 /* delegation and replication info */
250 int mds; /* -1 if same authority as parent */
251 int ndist; /* >0 if replicated */
252 int dist[CEPH_MAX_DIRFRAG_REP];
253};
254
255/*
256 * We cache inode xattrs as an encoded blob until they are first used,
257 * at which point we parse them into an rbtree.
258 */
259struct ceph_inode_xattr {
260 struct rb_node node;
261
262 const char *name;
263 int name_len;
264 const char *val;
265 int val_len;
266 int dirty;
267
268 int should_free_name;
269 int should_free_val;
270};
271
272struct ceph_inode_xattrs_info {
273 /*
274 * (still encoded) xattr blob. we avoid the overhead of parsing
275 * this until someone actually calls getxattr, etc.
276 *
277 * blob->vec.iov_len == 4 implies there are no xattrs; blob ==
278 * NULL means we don't know.
279 */
280 struct ceph_buffer *blob, *prealloc_blob;
281
282 struct rb_root index;
283 bool dirty;
284 int count;
285 int names_size;
286 int vals_size;
287 u64 version, index_version;
288};
289
290/*
291 * Ceph inode.
292 */
293#define CEPH_I_COMPLETE 1 /* we have complete directory cached */
294#define CEPH_I_NODELAY 4 /* do not delay cap release */
295#define CEPH_I_FLUSH 8 /* do not delay flush of dirty metadata */
Sage Weile9964c12010-03-01 15:16:56 -0800296#define CEPH_I_NOFLUSH 16 /* do not flush dirty caps */
Sage Weilde576062009-10-06 11:31:07 -0700297
298struct ceph_inode_info {
299 struct ceph_vino i_vino; /* ceph ino + snap */
300
301 u64 i_version;
302 u32 i_time_warp_seq;
303
304 unsigned i_ceph_flags;
305 unsigned long i_release_count;
306
307 struct ceph_file_layout i_layout;
308 char *i_symlink;
309
310 /* for dirs */
311 struct timespec i_rctime;
312 u64 i_rbytes, i_rfiles, i_rsubdirs;
313 u64 i_files, i_subdirs;
314 u64 i_max_offset; /* largest readdir offset, set with I_COMPLETE */
315
316 struct rb_root i_fragtree;
317 struct mutex i_fragtree_mutex;
318
319 struct ceph_inode_xattrs_info i_xattrs;
320
321 /* capabilities. protected _both_ by i_lock and cap->session's
322 * s_mutex. */
323 struct rb_root i_caps; /* cap list */
324 struct ceph_cap *i_auth_cap; /* authoritative cap, if any */
325 unsigned i_dirty_caps, i_flushing_caps; /* mask of dirtied fields */
326 struct list_head i_dirty_item, i_flushing_item;
327 u64 i_cap_flush_seq;
328 /* we need to track cap writeback on a per-cap-bit basis, to allow
329 * overlapping, pipelined cap flushes to the mds. we can probably
330 * reduce the tid to 8 bits if we're concerned about inode size. */
331 u16 i_cap_flush_last_tid, i_cap_flush_tid[CEPH_CAP_BITS];
332 wait_queue_head_t i_cap_wq; /* threads waiting on a capability */
333 unsigned long i_hold_caps_min; /* jiffies */
334 unsigned long i_hold_caps_max; /* jiffies */
335 struct list_head i_cap_delay_list; /* for delayed cap release to mds */
336 int i_cap_exporting_mds; /* to handle cap migration between */
337 unsigned i_cap_exporting_mseq; /* mds's. */
338 unsigned i_cap_exporting_issued;
339 struct ceph_cap_reservation i_cap_migration_resv;
340 struct list_head i_cap_snaps; /* snapped state pending flush to mds */
341 struct ceph_snap_context *i_head_snapc; /* set if wr_buffer_head > 0 */
342 unsigned i_snap_caps; /* cap bits for snapped files */
343
344 int i_nr_by_mode[CEPH_FILE_MODE_NUM]; /* open file counts */
345
346 u32 i_truncate_seq; /* last truncate to smaller size */
347 u64 i_truncate_size; /* and the size we last truncated down to */
348 int i_truncate_pending; /* still need to call vmtruncate */
349
350 u64 i_max_size; /* max file size authorized by mds */
351 u64 i_reported_size; /* (max_)size reported to or requested of mds */
352 u64 i_wanted_max_size; /* offset we'd like to write too */
353 u64 i_requested_max_size; /* max_size we've requested */
354
355 /* held references to caps */
356 int i_pin_ref;
357 int i_rd_ref, i_rdcache_ref, i_wr_ref;
358 int i_wrbuffer_ref, i_wrbuffer_ref_head;
359 u32 i_shared_gen; /* increment each time we get FILE_SHARED */
360 u32 i_rdcache_gen; /* we increment this each time we get
361 FILE_CACHE. If it's non-zero, we
362 _may_ have cached pages. */
363 u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */
364
365 struct list_head i_unsafe_writes; /* uncommitted sync writes */
366 struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */
367 spinlock_t i_unsafe_lock;
368
369 struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */
370 int i_snap_realm_counter; /* snap realm (if caps) */
371 struct list_head i_snap_realm_item;
372 struct list_head i_snap_flush_item;
373
374 struct work_struct i_wb_work; /* writeback work */
375 struct work_struct i_pg_inv_work; /* page invalidation work */
376
377 struct work_struct i_vmtruncate_work;
378
379 struct inode vfs_inode; /* at end */
380};
381
382static inline struct ceph_inode_info *ceph_inode(struct inode *inode)
383{
Noah Watkinsfbbccec2009-10-28 11:54:49 -0700384 return container_of(inode, struct ceph_inode_info, vfs_inode);
Sage Weilde576062009-10-06 11:31:07 -0700385}
386
387static inline void ceph_i_clear(struct inode *inode, unsigned mask)
388{
389 struct ceph_inode_info *ci = ceph_inode(inode);
390
391 spin_lock(&inode->i_lock);
392 ci->i_ceph_flags &= ~mask;
393 spin_unlock(&inode->i_lock);
394}
395
396static inline void ceph_i_set(struct inode *inode, unsigned mask)
397{
398 struct ceph_inode_info *ci = ceph_inode(inode);
399
400 spin_lock(&inode->i_lock);
401 ci->i_ceph_flags |= mask;
402 spin_unlock(&inode->i_lock);
403}
404
405static inline bool ceph_i_test(struct inode *inode, unsigned mask)
406{
407 struct ceph_inode_info *ci = ceph_inode(inode);
408 bool r;
409
410 smp_mb();
411 r = (ci->i_ceph_flags & mask) == mask;
412 return r;
413}
414
415
416/* find a specific frag @f */
417extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci,
418 u32 f);
419
420/*
421 * choose fragment for value @v. copy frag content to pfrag, if leaf
422 * exists
423 */
424extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v,
425 struct ceph_inode_frag *pfrag,
426 int *found);
427
428/*
429 * Ceph dentry state
430 */
431struct ceph_dentry_info {
432 struct ceph_mds_session *lease_session;
433 u32 lease_gen, lease_shared_gen;
434 u32 lease_seq;
435 unsigned long lease_renew_after, lease_renew_from;
436 struct list_head lru;
437 struct dentry *dentry;
438 u64 time;
439 u64 offset;
440};
441
442static inline struct ceph_dentry_info *ceph_dentry(struct dentry *dentry)
443{
444 return (struct ceph_dentry_info *)dentry->d_fsdata;
445}
446
447static inline loff_t ceph_make_fpos(unsigned frag, unsigned off)
448{
449 return ((loff_t)frag << 32) | (loff_t)off;
450}
451
452/*
453 * ino_t is <64 bits on many architectures, blech.
454 *
455 * don't include snap in ino hash, at least for now.
456 */
457static inline ino_t ceph_vino_to_ino(struct ceph_vino vino)
458{
459 ino_t ino = (ino_t)vino.ino; /* ^ (vino.snap << 20); */
460#if BITS_PER_LONG == 32
461 ino ^= vino.ino >> (sizeof(u64)-sizeof(ino_t)) * 8;
462 if (!ino)
463 ino = 1;
464#endif
465 return ino;
466}
467
468static inline int ceph_set_ino_cb(struct inode *inode, void *data)
469{
470 ceph_inode(inode)->i_vino = *(struct ceph_vino *)data;
471 inode->i_ino = ceph_vino_to_ino(*(struct ceph_vino *)data);
472 return 0;
473}
474
475static inline struct ceph_vino ceph_vino(struct inode *inode)
476{
477 return ceph_inode(inode)->i_vino;
478}
479
480/* for printf-style formatting */
481#define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap
482
483static inline u64 ceph_ino(struct inode *inode)
484{
485 return ceph_inode(inode)->i_vino.ino;
486}
487static inline u64 ceph_snap(struct inode *inode)
488{
489 return ceph_inode(inode)->i_vino.snap;
490}
491
492static inline int ceph_ino_compare(struct inode *inode, void *data)
493{
494 struct ceph_vino *pvino = (struct ceph_vino *)data;
495 struct ceph_inode_info *ci = ceph_inode(inode);
496 return ci->i_vino.ino == pvino->ino &&
497 ci->i_vino.snap == pvino->snap;
498}
499
500static inline struct inode *ceph_find_inode(struct super_block *sb,
501 struct ceph_vino vino)
502{
503 ino_t t = ceph_vino_to_ino(vino);
504 return ilookup5(sb, t, ceph_ino_compare, &vino);
505}
506
507
508/*
509 * caps helpers
510 */
511static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci)
512{
513 return !RB_EMPTY_ROOT(&ci->i_caps);
514}
515
516extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented);
517extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t);
518extern int __ceph_caps_issued_other(struct ceph_inode_info *ci,
519 struct ceph_cap *cap);
520
521static inline int ceph_caps_issued(struct ceph_inode_info *ci)
522{
523 int issued;
524 spin_lock(&ci->vfs_inode.i_lock);
525 issued = __ceph_caps_issued(ci, NULL);
526 spin_unlock(&ci->vfs_inode.i_lock);
527 return issued;
528}
529
530static inline int ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask,
531 int touch)
532{
533 int r;
534 spin_lock(&ci->vfs_inode.i_lock);
535 r = __ceph_caps_issued_mask(ci, mask, touch);
536 spin_unlock(&ci->vfs_inode.i_lock);
537 return r;
538}
539
540static inline int __ceph_caps_dirty(struct ceph_inode_info *ci)
541{
542 return ci->i_dirty_caps | ci->i_flushing_caps;
543}
Sage Weilafcdaea2009-10-14 14:27:38 -0700544extern void __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask);
Sage Weilde576062009-10-06 11:31:07 -0700545
546extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask);
547extern int __ceph_caps_used(struct ceph_inode_info *ci);
548
549extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci);
550
551/*
552 * wanted, by virtue of open file modes AND cap refs (buffered/cached data)
553 */
554static inline int __ceph_caps_wanted(struct ceph_inode_info *ci)
555{
556 int w = __ceph_caps_file_wanted(ci) | __ceph_caps_used(ci);
557 if (w & CEPH_CAP_FILE_BUFFER)
558 w |= CEPH_CAP_FILE_EXCL; /* we want EXCL if dirty data */
559 return w;
560}
561
562/* what the mds thinks we want */
563extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci);
564
565extern void ceph_caps_init(void);
566extern void ceph_caps_finalize(void);
Sage Weil85ccce42010-02-17 10:02:43 -0800567extern void ceph_adjust_min_caps(int delta);
Sage Weilde576062009-10-06 11:31:07 -0700568extern int ceph_reserve_caps(struct ceph_cap_reservation *ctx, int need);
569extern int ceph_unreserve_caps(struct ceph_cap_reservation *ctx);
570extern void ceph_reservation_status(struct ceph_client *client,
571 int *total, int *avail, int *used,
Sage Weil85ccce42010-02-17 10:02:43 -0800572 int *reserved, int *min);
Sage Weilde576062009-10-06 11:31:07 -0700573
574static inline struct ceph_client *ceph_inode_to_client(struct inode *inode)
575{
576 return (struct ceph_client *)inode->i_sb->s_fs_info;
577}
578
579static inline struct ceph_client *ceph_sb_to_client(struct super_block *sb)
580{
581 return (struct ceph_client *)sb->s_fs_info;
582}
583
Sage Weilde576062009-10-06 11:31:07 -0700584
585/*
586 * we keep buffered readdir results attached to file->private_data
587 */
588struct ceph_file_info {
589 int fmode; /* initialized on open */
590
591 /* readdir: position within the dir */
592 u32 frag;
593 struct ceph_mds_request *last_readdir;
594 int at_end;
595
596 /* readdir: position within a frag */
597 unsigned offset; /* offset of last chunk, adjusted for . and .. */
598 u64 next_offset; /* offset of next chunk (last_name's + 1) */
599 char *last_name; /* last entry in previous chunk */
600 struct dentry *dentry; /* next dentry (for dcache readdir) */
601 unsigned long dir_release_count;
602
603 /* used for -o dirstat read() on directory thing */
604 char *dir_info;
605 int dir_info_len;
606};
607
608
609
610/*
611 * snapshots
612 */
613
614/*
615 * A "snap context" is the set of existing snapshots when we
616 * write data. It is used by the OSD to guide its COW behavior.
617 *
618 * The ceph_snap_context is refcounted, and attached to each dirty
619 * page, indicating which context the dirty data belonged when it was
620 * dirtied.
621 */
622struct ceph_snap_context {
623 atomic_t nref;
624 u64 seq;
625 int num_snaps;
626 u64 snaps[];
627};
628
629static inline struct ceph_snap_context *
630ceph_get_snap_context(struct ceph_snap_context *sc)
631{
632 /*
633 printk("get_snap_context %p %d -> %d\n", sc, atomic_read(&sc->nref),
634 atomic_read(&sc->nref)+1);
635 */
636 if (sc)
637 atomic_inc(&sc->nref);
638 return sc;
639}
640
641static inline void ceph_put_snap_context(struct ceph_snap_context *sc)
642{
643 if (!sc)
644 return;
645 /*
646 printk("put_snap_context %p %d -> %d\n", sc, atomic_read(&sc->nref),
647 atomic_read(&sc->nref)-1);
648 */
649 if (atomic_dec_and_test(&sc->nref)) {
650 /*printk(" deleting snap_context %p\n", sc);*/
651 kfree(sc);
652 }
653}
654
655/*
656 * A "snap realm" describes a subset of the file hierarchy sharing
657 * the same set of snapshots that apply to it. The realms themselves
658 * are organized into a hierarchy, such that children inherit (some of)
659 * the snapshots of their parents.
660 *
661 * All inodes within the realm that have capabilities are linked into a
662 * per-realm list.
663 */
664struct ceph_snap_realm {
665 u64 ino;
666 atomic_t nref;
Sage Weila105f002010-02-15 14:37:55 -0800667 struct rb_node node;
668
Sage Weilde576062009-10-06 11:31:07 -0700669 u64 created, seq;
670 u64 parent_ino;
671 u64 parent_since; /* snapid when our current parent became so */
672
673 u64 *prior_parent_snaps; /* snaps inherited from any parents we */
674 int num_prior_parent_snaps; /* had prior to parent_since */
675 u64 *snaps; /* snaps specific to this realm */
676 int num_snaps;
677
678 struct ceph_snap_realm *parent;
679 struct list_head children; /* list of child realms */
680 struct list_head child_item;
681
682 struct list_head empty_item; /* if i have ref==0 */
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
691
692
693/*
694 * calculate the number of pages a given length and offset map onto,
695 * if we align the data.
696 */
697static inline int calc_pages_for(u64 off, u64 len)
698{
699 return ((off+len+PAGE_CACHE_SIZE-1) >> PAGE_CACHE_SHIFT) -
700 (off >> PAGE_CACHE_SHIFT);
701}
702
703
704
705/* snap.c */
706struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc,
707 u64 ino);
708extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc,
709 struct ceph_snap_realm *realm);
710extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc,
711 struct ceph_snap_realm *realm);
712extern int ceph_update_snap_trace(struct ceph_mds_client *m,
713 void *p, void *e, bool deletion);
714extern void ceph_handle_snap(struct ceph_mds_client *mdsc,
Sage Weil2600d2d2010-02-22 15:12:16 -0800715 struct ceph_mds_session *session,
Sage Weilde576062009-10-06 11:31:07 -0700716 struct ceph_msg *msg);
Sage Weilfc837c8f2010-04-13 11:41:22 -0700717extern void ceph_queue_cap_snap(struct ceph_inode_info *ci);
Sage Weilde576062009-10-06 11:31:07 -0700718extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci,
719 struct ceph_cap_snap *capsnap);
720extern void ceph_cleanup_empty_realms(struct ceph_mds_client *mdsc);
721
722/*
723 * a cap_snap is "pending" if it is still awaiting an in-progress
724 * sync write (that may/may not still update size, mtime, etc.).
725 */
726static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci)
727{
728 return !list_empty(&ci->i_cap_snaps) &&
729 list_entry(ci->i_cap_snaps.prev, struct ceph_cap_snap,
730 ci_item)->writing;
731}
732
733
734/* super.c */
735extern struct kmem_cache *ceph_inode_cachep;
736extern struct kmem_cache *ceph_cap_cachep;
737extern struct kmem_cache *ceph_dentry_cachep;
738extern struct kmem_cache *ceph_file_cachep;
739
740extern const char *ceph_msg_type_name(int type);
Sage Weil07433042009-11-18 16:50:41 -0800741extern int ceph_check_fsid(struct ceph_client *client, struct ceph_fsid *fsid);
Sage Weilde576062009-10-06 11:31:07 -0700742
743#define FSID_FORMAT "%02x%02x%02x%02x-%02x%02x-%02x%02x-%02x%02x-" \
744 "%02x%02x%02x%02x%02x%02x"
745#define PR_FSID(f) (f)->fsid[0], (f)->fsid[1], (f)->fsid[2], (f)->fsid[3], \
746 (f)->fsid[4], (f)->fsid[5], (f)->fsid[6], (f)->fsid[7], \
747 (f)->fsid[8], (f)->fsid[9], (f)->fsid[10], (f)->fsid[11], \
748 (f)->fsid[12], (f)->fsid[13], (f)->fsid[14], (f)->fsid[15]
749
750/* inode.c */
751extern const struct inode_operations ceph_file_iops;
752
753extern struct inode *ceph_alloc_inode(struct super_block *sb);
754extern void ceph_destroy_inode(struct inode *inode);
755
756extern struct inode *ceph_get_inode(struct super_block *sb,
757 struct ceph_vino vino);
758extern struct inode *ceph_get_snapdir(struct inode *parent);
759extern int ceph_fill_file_size(struct inode *inode, int issued,
760 u32 truncate_seq, u64 truncate_size, u64 size);
761extern void ceph_fill_file_time(struct inode *inode, int issued,
762 u64 time_warp_seq, struct timespec *ctime,
763 struct timespec *mtime, struct timespec *atime);
764extern int ceph_fill_trace(struct super_block *sb,
765 struct ceph_mds_request *req,
766 struct ceph_mds_session *session);
767extern int ceph_readdir_prepopulate(struct ceph_mds_request *req,
768 struct ceph_mds_session *session);
769
770extern int ceph_inode_holds_cap(struct inode *inode, int mask);
771
772extern int ceph_inode_set_size(struct inode *inode, loff_t size);
Sage Weilde576062009-10-06 11:31:07 -0700773extern void __ceph_do_pending_vmtruncate(struct inode *inode);
Sage Weil3c6f6b72010-02-09 15:24:44 -0800774extern void ceph_queue_vmtruncate(struct inode *inode);
775
776extern void ceph_queue_invalidate(struct inode *inode);
777extern void ceph_queue_writeback(struct inode *inode);
Sage Weilde576062009-10-06 11:31:07 -0700778
779extern int ceph_do_getattr(struct inode *inode, int mask);
780extern int ceph_permission(struct inode *inode, int mask);
781extern int ceph_setattr(struct dentry *dentry, struct iattr *attr);
782extern int ceph_getattr(struct vfsmount *mnt, struct dentry *dentry,
783 struct kstat *stat);
784
785/* xattr.c */
786extern int ceph_setxattr(struct dentry *, const char *, const void *,
787 size_t, int);
788extern ssize_t ceph_getxattr(struct dentry *, const char *, void *, size_t);
789extern ssize_t ceph_listxattr(struct dentry *, char *, size_t);
790extern int ceph_removexattr(struct dentry *, const char *);
791extern void __ceph_build_xattrs_blob(struct ceph_inode_info *ci);
792extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci);
793
794/* caps.c */
795extern const char *ceph_cap_string(int c);
796extern void ceph_handle_caps(struct ceph_mds_session *session,
797 struct ceph_msg *msg);
798extern int ceph_add_cap(struct inode *inode,
799 struct ceph_mds_session *session, u64 cap_id,
800 int fmode, unsigned issued, unsigned wanted,
801 unsigned cap, unsigned seq, u64 realmino, int flags,
802 struct ceph_cap_reservation *caps_reservation);
Sage Weil7c1332b2010-02-16 11:39:45 -0800803extern void __ceph_remove_cap(struct ceph_cap *cap);
Sage Weilde576062009-10-06 11:31:07 -0700804static inline void ceph_remove_cap(struct ceph_cap *cap)
805{
806 struct inode *inode = &cap->ci->vfs_inode;
807 spin_lock(&inode->i_lock);
Sage Weil7c1332b2010-02-16 11:39:45 -0800808 __ceph_remove_cap(cap);
Sage Weilde576062009-10-06 11:31:07 -0700809 spin_unlock(&inode->i_lock);
810}
Sage Weil7c1332b2010-02-16 11:39:45 -0800811extern void ceph_put_cap(struct ceph_cap *cap);
Sage Weilde576062009-10-06 11:31:07 -0700812
813extern void ceph_queue_caps_release(struct inode *inode);
Stephen Rothwellf1a3d572010-01-18 11:53:08 +1100814extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc);
Sage Weilde576062009-10-06 11:31:07 -0700815extern int ceph_fsync(struct file *file, struct dentry *dentry, int datasync);
816extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc,
817 struct ceph_mds_session *session);
818extern int ceph_get_cap_mds(struct inode *inode);
819extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps);
820extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had);
821extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
822 struct ceph_snap_context *snapc);
823extern void __ceph_flush_snaps(struct ceph_inode_info *ci,
824 struct ceph_mds_session **psession);
825extern void ceph_check_caps(struct ceph_inode_info *ci, int flags,
826 struct ceph_mds_session *session);
Sage Weilafcdaea2009-10-14 14:27:38 -0700827extern void ceph_check_delayed_caps(struct ceph_mds_client *mdsc);
828extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc);
Sage Weilde576062009-10-06 11:31:07 -0700829
830extern int ceph_encode_inode_release(void **p, struct inode *inode,
831 int mds, int drop, int unless, int force);
832extern int ceph_encode_dentry_release(void **p, struct dentry *dn,
833 int mds, int drop, int unless);
834
835extern int ceph_get_caps(struct ceph_inode_info *ci, int need, int want,
836 int *got, loff_t endoff);
837
838/* for counting open files by mode */
839static inline void __ceph_get_fmode(struct ceph_inode_info *ci, int mode)
840{
841 ci->i_nr_by_mode[mode]++;
842}
843extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode);
844
845/* addr.c */
846extern const struct address_space_operations ceph_aops;
847extern int ceph_mmap(struct file *file, struct vm_area_struct *vma);
848
849/* file.c */
850extern const struct file_operations ceph_file_fops;
851extern const struct address_space_operations ceph_aops;
852extern int ceph_open(struct inode *inode, struct file *file);
853extern struct dentry *ceph_lookup_open(struct inode *dir, struct dentry *dentry,
854 struct nameidata *nd, int mode,
855 int locked_dir);
856extern int ceph_release(struct inode *inode, struct file *filp);
857extern void ceph_release_page_vector(struct page **pages, int num_pages);
858
859/* dir.c */
860extern const struct file_operations ceph_dir_fops;
861extern const struct inode_operations ceph_dir_iops;
862extern struct dentry_operations ceph_dentry_ops, ceph_snap_dentry_ops,
863 ceph_snapdir_dentry_ops;
864
865extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry);
866extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req,
867 struct dentry *dentry, int err);
868
869extern void ceph_dentry_lru_add(struct dentry *dn);
870extern void ceph_dentry_lru_touch(struct dentry *dn);
871extern void ceph_dentry_lru_del(struct dentry *dn);
872
873/*
874 * our d_ops vary depending on whether the inode is live,
875 * snapshotted (read-only), or a virtual ".snap" directory.
876 */
877int ceph_init_dentry(struct dentry *dentry);
878
879
880/* ioctl.c */
881extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
882
883/* export.c */
884extern const struct export_operations ceph_export_ops;
885
886/* debugfs.c */
887extern int ceph_debugfs_init(void);
888extern void ceph_debugfs_cleanup(void);
889extern int ceph_debugfs_client_init(struct ceph_client *client);
890extern void ceph_debugfs_client_cleanup(struct ceph_client *client);
891
892static inline struct inode *get_dentry_parent_inode(struct dentry *dentry)
893{
894 if (dentry && dentry->d_parent)
895 return dentry->d_parent->d_inode;
896
897 return NULL;
898}
899
900#endif /* _FS_CEPH_SUPER_H */