blob: 1886294e12f7a3106c00f3376d92e43330514081 [file] [log] [blame]
Sage Weilde576062009-10-06 11:31:07 -07001#ifndef _FS_CEPH_SUPER_H
2#define _FS_CEPH_SUPER_H
3
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -07004#include <linux/ceph/ceph_debug.h>
Sage Weilde576062009-10-06 11:31:07 -07005
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>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090015#include <linux/slab.h>
Sage Weilde576062009-10-06 11:31:07 -070016
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070017#include <linux/ceph/libceph.h>
Sage Weilde576062009-10-06 11:31:07 -070018
19/* f_type in struct statfs */
20#define CEPH_SUPER_MAGIC 0x00c36400
21
22/* large granularity for statfs utilization stats to facilitate
23 * large volume sizes on 32-bit machines. */
24#define CEPH_BLOCK_SHIFT 20 /* 1 MB */
25#define CEPH_BLOCK (1 << CEPH_BLOCK_SHIFT)
26
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070027#define CEPH_MOUNT_OPT_DIRSTAT (1<<4) /* `cat dirname` for stats */
28#define CEPH_MOUNT_OPT_RBYTES (1<<5) /* dir st_bytes = rbytes */
29#define CEPH_MOUNT_OPT_NOASYNCREADDIR (1<<7) /* no dcache readdir */
Sage Weil6a259382010-07-07 09:06:08 -070030
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070031#define CEPH_MOUNT_OPT_DEFAULT (CEPH_MOUNT_OPT_RBYTES)
Sage Weilde576062009-10-06 11:31:07 -070032
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070033#define ceph_set_mount_opt(fsc, opt) \
34 (fsc)->mount_options->flags |= CEPH_MOUNT_OPT_##opt;
35#define ceph_test_mount_opt(fsc, opt) \
36 (!!((fsc)->mount_options->flags & CEPH_MOUNT_OPT_##opt))
Sage Weilde576062009-10-06 11:31:07 -070037
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070038#define CEPH_MAX_READDIR_DEFAULT 1024
39#define CEPH_MAX_READDIR_BYTES_DEFAULT (512*1024)
40#define CEPH_SNAPDIRNAME_DEFAULT ".snap"
Sage Weilde576062009-10-06 11:31:07 -070041
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070042struct ceph_mount_options {
Sage Weilde576062009-10-06 11:31:07 -070043 int flags;
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070044 int sb_flags;
45
Sage Weil6e19a162010-04-29 16:38:32 -070046 int wsize;
47 int rsize; /* max readahead */
48 int congestion_kb; /* max writeback in flight */
49 int caps_wanted_delay_min, caps_wanted_delay_max;
50 int cap_release_safety;
Sage Weil23804d92010-05-14 13:06:30 -070051 int max_readdir; /* max readdir result (entires) */
52 int max_readdir_bytes; /* max readdir result (bytes) */
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070053
54 /*
55 * everything above this point can be memcmp'd; everything below
56 * is handled in compare_mount_options()
57 */
58
Sage Weilde576062009-10-06 11:31:07 -070059 char *snapdir_name; /* default ".snap" */
Sage Weilde576062009-10-06 11:31:07 -070060};
61
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070062struct ceph_fs_client {
Sage Weilde576062009-10-06 11:31:07 -070063 struct super_block *sb;
64
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070065 struct ceph_mount_options *mount_options;
66 struct ceph_client *client;
67
Sage Weilde576062009-10-06 11:31:07 -070068 unsigned long mount_state;
Sage Weil85ccce42010-02-17 10:02:43 -080069 int min_caps; /* min caps i added */
70
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070071 struct ceph_mds_client *mdsc;
Sage Weilde576062009-10-06 11:31:07 -070072
73 /* writeback */
74 mempool_t *wb_pagevec_pool;
75 struct workqueue_struct *wb_wq;
76 struct workqueue_struct *pg_inv_wq;
77 struct workqueue_struct *trunc_wq;
Yehuda Sadeh2baba252009-12-18 13:51:57 -080078 atomic_long_t writeback_count;
Sage Weilde576062009-10-06 11:31:07 -070079
80 struct backing_dev_info backing_dev_info;
Sage Weil07433042009-11-18 16:50:41 -080081
82#ifdef CONFIG_DEBUG_FS
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070083 struct dentry *debugfs_dentry_lru, *debugfs_caps;
Yehuda Sadeh2baba252009-12-18 13:51:57 -080084 struct dentry *debugfs_congestion_kb;
Sage Weil06edf042009-12-15 14:44:32 -080085 struct dentry *debugfs_bdi;
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070086 struct dentry *debugfs_mdsc, *debugfs_mdsmap;
Sage Weil07433042009-11-18 16:50:41 -080087#endif
Sage Weilde576062009-10-06 11:31:07 -070088};
89
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070090
Sage Weilde576062009-10-06 11:31:07 -070091/*
92 * File i/o capability. This tracks shared state with the metadata
93 * server that allows us to cache or writeback attributes or to read
94 * and write data. For any given inode, we should have one or more
95 * capabilities, one issued by each metadata server, and our
96 * cumulative access is the OR of all issued capabilities.
97 *
98 * Each cap is referenced by the inode's i_caps rbtree and by per-mds
99 * session capability lists.
100 */
101struct ceph_cap {
102 struct ceph_inode_info *ci;
103 struct rb_node ci_node; /* per-ci cap tree */
104 struct ceph_mds_session *session;
105 struct list_head session_caps; /* per-session caplist */
106 int mds;
107 u64 cap_id; /* unique cap id (mds provided) */
108 int issued; /* latest, from the mds */
109 int implemented; /* implemented superset of issued (for revocation) */
110 int mds_wanted;
Sage Weil685f9a5d2009-11-09 12:05:48 -0800111 u32 seq, issue_seq, mseq;
112 u32 cap_gen; /* active/stale cycle */
Sage Weilde576062009-10-06 11:31:07 -0700113 unsigned long last_used;
114 struct list_head caps_item;
115};
116
117#define CHECK_CAPS_NODELAY 1 /* do not delay any further */
118#define CHECK_CAPS_AUTHONLY 2 /* only check auth cap */
119#define CHECK_CAPS_FLUSH 4 /* flush any dirty caps */
120
121/*
122 * Snapped cap state that is pending flush to mds. When a snapshot occurs,
123 * we first complete any in-process sync writes and writeback any dirty
124 * data before flushing the snapped state (tracked here) back to the MDS.
125 */
126struct ceph_cap_snap {
127 atomic_t nref;
128 struct ceph_inode_info *ci;
129 struct list_head ci_item, flushing_item;
130
131 u64 follows, flush_tid;
132 int issued, dirty;
133 struct ceph_snap_context *context;
134
135 mode_t mode;
136 uid_t uid;
137 gid_t gid;
138
Sage Weil4a625be2010-08-22 15:03:56 -0700139 struct ceph_buffer *xattr_blob;
Sage Weilde576062009-10-06 11:31:07 -0700140 u64 xattr_version;
141
142 u64 size;
143 struct timespec mtime, atime, ctime;
144 u64 time_warp_seq;
145 int writing; /* a sync write is still in progress */
146 int dirty_pages; /* dirty pages awaiting writeback */
147};
148
149static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap)
150{
Sage Weil4a625be2010-08-22 15:03:56 -0700151 if (atomic_dec_and_test(&capsnap->nref)) {
152 if (capsnap->xattr_blob)
153 ceph_buffer_put(capsnap->xattr_blob);
Sage Weilde576062009-10-06 11:31:07 -0700154 kfree(capsnap);
Sage Weil4a625be2010-08-22 15:03:56 -0700155 }
Sage Weilde576062009-10-06 11:31:07 -0700156}
157
158/*
159 * The frag tree describes how a directory is fragmented, potentially across
160 * multiple metadata servers. It is also used to indicate points where
161 * metadata authority is delegated, and whether/where metadata is replicated.
162 *
163 * A _leaf_ frag will be present in the i_fragtree IFF there is
164 * delegation info. That is, if mds >= 0 || ndist > 0.
165 */
166#define CEPH_MAX_DIRFRAG_REP 4
167
168struct ceph_inode_frag {
169 struct rb_node node;
170
171 /* fragtree state */
172 u32 frag;
173 int split_by; /* i.e. 2^(split_by) children */
174
175 /* delegation and replication info */
176 int mds; /* -1 if same authority as parent */
177 int ndist; /* >0 if replicated */
178 int dist[CEPH_MAX_DIRFRAG_REP];
179};
180
181/*
182 * We cache inode xattrs as an encoded blob until they are first used,
183 * at which point we parse them into an rbtree.
184 */
185struct ceph_inode_xattr {
186 struct rb_node node;
187
188 const char *name;
189 int name_len;
190 const char *val;
191 int val_len;
192 int dirty;
193
194 int should_free_name;
195 int should_free_val;
196};
197
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700198/*
199 * Ceph dentry state
200 */
201struct ceph_dentry_info {
202 struct ceph_mds_session *lease_session;
203 u32 lease_gen, lease_shared_gen;
204 u32 lease_seq;
205 unsigned long lease_renew_after, lease_renew_from;
206 struct list_head lru;
207 struct dentry *dentry;
208 u64 time;
209 u64 offset;
210};
211
Sage Weilde576062009-10-06 11:31:07 -0700212struct ceph_inode_xattrs_info {
213 /*
214 * (still encoded) xattr blob. we avoid the overhead of parsing
215 * this until someone actually calls getxattr, etc.
216 *
217 * blob->vec.iov_len == 4 implies there are no xattrs; blob ==
218 * NULL means we don't know.
219 */
220 struct ceph_buffer *blob, *prealloc_blob;
221
222 struct rb_root index;
223 bool dirty;
224 int count;
225 int names_size;
226 int vals_size;
227 u64 version, index_version;
228};
229
230/*
231 * Ceph inode.
232 */
Sage Weilde576062009-10-06 11:31:07 -0700233struct ceph_inode_info {
234 struct ceph_vino i_vino; /* ceph ino + snap */
235
236 u64 i_version;
237 u32 i_time_warp_seq;
238
239 unsigned i_ceph_flags;
240 unsigned long i_release_count;
241
242 struct ceph_file_layout i_layout;
243 char *i_symlink;
244
245 /* for dirs */
246 struct timespec i_rctime;
247 u64 i_rbytes, i_rfiles, i_rsubdirs;
248 u64 i_files, i_subdirs;
249 u64 i_max_offset; /* largest readdir offset, set with I_COMPLETE */
250
251 struct rb_root i_fragtree;
252 struct mutex i_fragtree_mutex;
253
254 struct ceph_inode_xattrs_info i_xattrs;
255
256 /* capabilities. protected _both_ by i_lock and cap->session's
257 * s_mutex. */
258 struct rb_root i_caps; /* cap list */
259 struct ceph_cap *i_auth_cap; /* authoritative cap, if any */
260 unsigned i_dirty_caps, i_flushing_caps; /* mask of dirtied fields */
261 struct list_head i_dirty_item, i_flushing_item;
262 u64 i_cap_flush_seq;
263 /* we need to track cap writeback on a per-cap-bit basis, to allow
264 * overlapping, pipelined cap flushes to the mds. we can probably
265 * reduce the tid to 8 bits if we're concerned about inode size. */
266 u16 i_cap_flush_last_tid, i_cap_flush_tid[CEPH_CAP_BITS];
267 wait_queue_head_t i_cap_wq; /* threads waiting on a capability */
268 unsigned long i_hold_caps_min; /* jiffies */
269 unsigned long i_hold_caps_max; /* jiffies */
270 struct list_head i_cap_delay_list; /* for delayed cap release to mds */
271 int i_cap_exporting_mds; /* to handle cap migration between */
272 unsigned i_cap_exporting_mseq; /* mds's. */
273 unsigned i_cap_exporting_issued;
274 struct ceph_cap_reservation i_cap_migration_resv;
275 struct list_head i_cap_snaps; /* snapped state pending flush to mds */
Sage Weil7d8cb262010-08-24 08:44:16 -0700276 struct ceph_snap_context *i_head_snapc; /* set if wr_buffer_head > 0 or
277 dirty|flushing caps */
Sage Weilde576062009-10-06 11:31:07 -0700278 unsigned i_snap_caps; /* cap bits for snapped files */
279
280 int i_nr_by_mode[CEPH_FILE_MODE_NUM]; /* open file counts */
281
282 u32 i_truncate_seq; /* last truncate to smaller size */
283 u64 i_truncate_size; /* and the size we last truncated down to */
284 int i_truncate_pending; /* still need to call vmtruncate */
285
286 u64 i_max_size; /* max file size authorized by mds */
287 u64 i_reported_size; /* (max_)size reported to or requested of mds */
288 u64 i_wanted_max_size; /* offset we'd like to write too */
289 u64 i_requested_max_size; /* max_size we've requested */
290
291 /* held references to caps */
292 int i_pin_ref;
293 int i_rd_ref, i_rdcache_ref, i_wr_ref;
294 int i_wrbuffer_ref, i_wrbuffer_ref_head;
295 u32 i_shared_gen; /* increment each time we get FILE_SHARED */
296 u32 i_rdcache_gen; /* we increment this each time we get
297 FILE_CACHE. If it's non-zero, we
298 _may_ have cached pages. */
299 u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */
300
301 struct list_head i_unsafe_writes; /* uncommitted sync writes */
302 struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */
303 spinlock_t i_unsafe_lock;
304
305 struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */
306 int i_snap_realm_counter; /* snap realm (if caps) */
307 struct list_head i_snap_realm_item;
308 struct list_head i_snap_flush_item;
309
310 struct work_struct i_wb_work; /* writeback work */
311 struct work_struct i_pg_inv_work; /* page invalidation work */
312
313 struct work_struct i_vmtruncate_work;
314
315 struct inode vfs_inode; /* at end */
316};
317
318static inline struct ceph_inode_info *ceph_inode(struct inode *inode)
319{
Noah Watkinsfbbccec2009-10-28 11:54:49 -0700320 return container_of(inode, struct ceph_inode_info, vfs_inode);
Sage Weilde576062009-10-06 11:31:07 -0700321}
322
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700323static inline struct ceph_vino ceph_vino(struct inode *inode)
324{
325 return ceph_inode(inode)->i_vino;
326}
327
328/*
329 * ino_t is <64 bits on many architectures, blech.
330 *
331 * don't include snap in ino hash, at least for now.
332 */
333static inline ino_t ceph_vino_to_ino(struct ceph_vino vino)
334{
335 ino_t ino = (ino_t)vino.ino; /* ^ (vino.snap << 20); */
336#if BITS_PER_LONG == 32
337 ino ^= vino.ino >> (sizeof(u64)-sizeof(ino_t)) * 8;
338 if (!ino)
339 ino = 1;
340#endif
341 return ino;
342}
343
344/* for printf-style formatting */
345#define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap
346
347static inline u64 ceph_ino(struct inode *inode)
348{
349 return ceph_inode(inode)->i_vino.ino;
350}
351static inline u64 ceph_snap(struct inode *inode)
352{
353 return ceph_inode(inode)->i_vino.snap;
354}
355
356static inline int ceph_ino_compare(struct inode *inode, void *data)
357{
358 struct ceph_vino *pvino = (struct ceph_vino *)data;
359 struct ceph_inode_info *ci = ceph_inode(inode);
360 return ci->i_vino.ino == pvino->ino &&
361 ci->i_vino.snap == pvino->snap;
362}
363
364static inline struct inode *ceph_find_inode(struct super_block *sb,
365 struct ceph_vino vino)
366{
367 ino_t t = ceph_vino_to_ino(vino);
368 return ilookup5(sb, t, ceph_ino_compare, &vino);
369}
370
371
372/*
373 * Ceph inode.
374 */
375#define CEPH_I_COMPLETE 1 /* we have complete directory cached */
376#define CEPH_I_NODELAY 4 /* do not delay cap release */
377#define CEPH_I_FLUSH 8 /* do not delay flush of dirty metadata */
378#define CEPH_I_NOFLUSH 16 /* do not flush dirty caps */
379
Sage Weilde576062009-10-06 11:31:07 -0700380static inline void ceph_i_clear(struct inode *inode, unsigned mask)
381{
382 struct ceph_inode_info *ci = ceph_inode(inode);
383
384 spin_lock(&inode->i_lock);
385 ci->i_ceph_flags &= ~mask;
386 spin_unlock(&inode->i_lock);
387}
388
389static inline void ceph_i_set(struct inode *inode, unsigned mask)
390{
391 struct ceph_inode_info *ci = ceph_inode(inode);
392
393 spin_lock(&inode->i_lock);
394 ci->i_ceph_flags |= mask;
395 spin_unlock(&inode->i_lock);
396}
397
398static inline bool ceph_i_test(struct inode *inode, unsigned mask)
399{
400 struct ceph_inode_info *ci = ceph_inode(inode);
401 bool r;
402
Sage Weilefa4c122010-10-18 14:04:31 -0700403 spin_lock(&inode->i_lock);
Sage Weilde576062009-10-06 11:31:07 -0700404 r = (ci->i_ceph_flags & mask) == mask;
Sage Weilefa4c122010-10-18 14:04:31 -0700405 spin_unlock(&inode->i_lock);
Sage Weilde576062009-10-06 11:31:07 -0700406 return r;
407}
408
409
410/* find a specific frag @f */
411extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci,
412 u32 f);
413
414/*
415 * choose fragment for value @v. copy frag content to pfrag, if leaf
416 * exists
417 */
418extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v,
419 struct ceph_inode_frag *pfrag,
420 int *found);
421
Sage Weilde576062009-10-06 11:31:07 -0700422static inline struct ceph_dentry_info *ceph_dentry(struct dentry *dentry)
423{
424 return (struct ceph_dentry_info *)dentry->d_fsdata;
425}
426
427static inline loff_t ceph_make_fpos(unsigned frag, unsigned off)
428{
429 return ((loff_t)frag << 32) | (loff_t)off;
430}
431
Sage Weilde576062009-10-06 11:31:07 -0700432static inline int ceph_set_ino_cb(struct inode *inode, void *data)
433{
434 ceph_inode(inode)->i_vino = *(struct ceph_vino *)data;
435 inode->i_ino = ceph_vino_to_ino(*(struct ceph_vino *)data);
436 return 0;
437}
438
Sage Weilde576062009-10-06 11:31:07 -0700439/*
440 * caps helpers
441 */
442static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci)
443{
444 return !RB_EMPTY_ROOT(&ci->i_caps);
445}
446
447extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented);
448extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t);
449extern int __ceph_caps_issued_other(struct ceph_inode_info *ci,
450 struct ceph_cap *cap);
451
452static inline int ceph_caps_issued(struct ceph_inode_info *ci)
453{
454 int issued;
455 spin_lock(&ci->vfs_inode.i_lock);
456 issued = __ceph_caps_issued(ci, NULL);
457 spin_unlock(&ci->vfs_inode.i_lock);
458 return issued;
459}
460
461static inline int ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask,
462 int touch)
463{
464 int r;
465 spin_lock(&ci->vfs_inode.i_lock);
466 r = __ceph_caps_issued_mask(ci, mask, touch);
467 spin_unlock(&ci->vfs_inode.i_lock);
468 return r;
469}
470
471static inline int __ceph_caps_dirty(struct ceph_inode_info *ci)
472{
473 return ci->i_dirty_caps | ci->i_flushing_caps;
474}
Sage Weilafcdaea2009-10-14 14:27:38 -0700475extern void __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask);
Sage Weilde576062009-10-06 11:31:07 -0700476
477extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask);
478extern int __ceph_caps_used(struct ceph_inode_info *ci);
479
480extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci);
481
482/*
483 * wanted, by virtue of open file modes AND cap refs (buffered/cached data)
484 */
485static inline int __ceph_caps_wanted(struct ceph_inode_info *ci)
486{
487 int w = __ceph_caps_file_wanted(ci) | __ceph_caps_used(ci);
488 if (w & CEPH_CAP_FILE_BUFFER)
489 w |= CEPH_CAP_FILE_EXCL; /* we want EXCL if dirty data */
490 return w;
491}
492
493/* what the mds thinks we want */
494extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci);
495
Yehuda Sadeh37151662010-06-17 16:16:12 -0700496extern void ceph_caps_init(struct ceph_mds_client *mdsc);
497extern void ceph_caps_finalize(struct ceph_mds_client *mdsc);
498extern void ceph_adjust_min_caps(struct ceph_mds_client *mdsc, int delta);
499extern int ceph_reserve_caps(struct ceph_mds_client *mdsc,
500 struct ceph_cap_reservation *ctx, int need);
501extern int ceph_unreserve_caps(struct ceph_mds_client *mdsc,
502 struct ceph_cap_reservation *ctx);
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700503extern void ceph_reservation_status(struct ceph_fs_client *client,
Sage Weilde576062009-10-06 11:31:07 -0700504 int *total, int *avail, int *used,
Sage Weil85ccce42010-02-17 10:02:43 -0800505 int *reserved, int *min);
Sage Weilde576062009-10-06 11:31:07 -0700506
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700507static inline struct ceph_fs_client *ceph_inode_to_client(struct inode *inode)
Sage Weilde576062009-10-06 11:31:07 -0700508{
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700509 return (struct ceph_fs_client *)inode->i_sb->s_fs_info;
Sage Weilde576062009-10-06 11:31:07 -0700510}
511
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700512static inline struct ceph_fs_client *ceph_sb_to_client(struct super_block *sb)
Sage Weilde576062009-10-06 11:31:07 -0700513{
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700514 return (struct ceph_fs_client *)sb->s_fs_info;
Sage Weilde576062009-10-06 11:31:07 -0700515}
516
Sage Weilde576062009-10-06 11:31:07 -0700517
518/*
519 * we keep buffered readdir results attached to file->private_data
520 */
521struct ceph_file_info {
522 int fmode; /* initialized on open */
523
524 /* readdir: position within the dir */
525 u32 frag;
526 struct ceph_mds_request *last_readdir;
527 int at_end;
528
529 /* readdir: position within a frag */
530 unsigned offset; /* offset of last chunk, adjusted for . and .. */
531 u64 next_offset; /* offset of next chunk (last_name's + 1) */
532 char *last_name; /* last entry in previous chunk */
533 struct dentry *dentry; /* next dentry (for dcache readdir) */
534 unsigned long dir_release_count;
535
536 /* used for -o dirstat read() on directory thing */
537 char *dir_info;
538 int dir_info_len;
539};
540
541
542
543/*
Sage Weilde576062009-10-06 11:31:07 -0700544 * A "snap realm" describes a subset of the file hierarchy sharing
545 * the same set of snapshots that apply to it. The realms themselves
546 * are organized into a hierarchy, such that children inherit (some of)
547 * the snapshots of their parents.
548 *
549 * All inodes within the realm that have capabilities are linked into a
550 * per-realm list.
551 */
552struct ceph_snap_realm {
553 u64 ino;
554 atomic_t nref;
Sage Weila105f002010-02-15 14:37:55 -0800555 struct rb_node node;
556
Sage Weilde576062009-10-06 11:31:07 -0700557 u64 created, seq;
558 u64 parent_ino;
559 u64 parent_since; /* snapid when our current parent became so */
560
561 u64 *prior_parent_snaps; /* snaps inherited from any parents we */
562 int num_prior_parent_snaps; /* had prior to parent_since */
563 u64 *snaps; /* snaps specific to this realm */
564 int num_snaps;
565
566 struct ceph_snap_realm *parent;
567 struct list_head children; /* list of child realms */
568 struct list_head child_item;
569
570 struct list_head empty_item; /* if i have ref==0 */
571
Sage Weilae00d4f2010-09-16 16:26:51 -0700572 struct list_head dirty_item; /* if realm needs new context */
573
Sage Weilde576062009-10-06 11:31:07 -0700574 /* the current set of snaps for this realm */
575 struct ceph_snap_context *cached_context;
576
577 struct list_head inodes_with_caps;
578 spinlock_t inodes_with_caps_lock;
579};
580
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700581static inline int default_congestion_kb(void)
Sage Weilde576062009-10-06 11:31:07 -0700582{
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700583 int congestion_kb;
584
585 /*
586 * Copied from NFS
587 *
588 * congestion size, scale with available memory.
589 *
590 * 64MB: 8192k
591 * 128MB: 11585k
592 * 256MB: 16384k
593 * 512MB: 23170k
594 * 1GB: 32768k
595 * 2GB: 46340k
596 * 4GB: 65536k
597 * 8GB: 92681k
598 * 16GB: 131072k
599 *
600 * This allows larger machines to have larger/more transfers.
601 * Limit the default to 256M
602 */
603 congestion_kb = (16*int_sqrt(totalram_pages)) << (PAGE_SHIFT-10);
604 if (congestion_kb > 256*1024)
605 congestion_kb = 256*1024;
606
607 return congestion_kb;
Sage Weilde576062009-10-06 11:31:07 -0700608}
609
610
611
612/* snap.c */
613struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc,
614 u64 ino);
615extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc,
616 struct ceph_snap_realm *realm);
617extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc,
618 struct ceph_snap_realm *realm);
619extern int ceph_update_snap_trace(struct ceph_mds_client *m,
620 void *p, void *e, bool deletion);
621extern void ceph_handle_snap(struct ceph_mds_client *mdsc,
Sage Weil2600d2d2010-02-22 15:12:16 -0800622 struct ceph_mds_session *session,
Sage Weilde576062009-10-06 11:31:07 -0700623 struct ceph_msg *msg);
Sage Weilfc837c8f2010-04-13 11:41:22 -0700624extern void ceph_queue_cap_snap(struct ceph_inode_info *ci);
Sage Weilde576062009-10-06 11:31:07 -0700625extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci,
626 struct ceph_cap_snap *capsnap);
627extern void ceph_cleanup_empty_realms(struct ceph_mds_client *mdsc);
628
629/*
630 * a cap_snap is "pending" if it is still awaiting an in-progress
631 * sync write (that may/may not still update size, mtime, etc.).
632 */
633static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci)
634{
635 return !list_empty(&ci->i_cap_snaps) &&
636 list_entry(ci->i_cap_snaps.prev, struct ceph_cap_snap,
637 ci_item)->writing;
638}
639
Sage Weilde576062009-10-06 11:31:07 -0700640/* inode.c */
641extern const struct inode_operations ceph_file_iops;
642
643extern struct inode *ceph_alloc_inode(struct super_block *sb);
644extern void ceph_destroy_inode(struct inode *inode);
645
646extern struct inode *ceph_get_inode(struct super_block *sb,
647 struct ceph_vino vino);
648extern struct inode *ceph_get_snapdir(struct inode *parent);
649extern int ceph_fill_file_size(struct inode *inode, int issued,
650 u32 truncate_seq, u64 truncate_size, u64 size);
651extern void ceph_fill_file_time(struct inode *inode, int issued,
652 u64 time_warp_seq, struct timespec *ctime,
653 struct timespec *mtime, struct timespec *atime);
654extern int ceph_fill_trace(struct super_block *sb,
655 struct ceph_mds_request *req,
656 struct ceph_mds_session *session);
657extern int ceph_readdir_prepopulate(struct ceph_mds_request *req,
658 struct ceph_mds_session *session);
659
660extern int ceph_inode_holds_cap(struct inode *inode, int mask);
661
662extern int ceph_inode_set_size(struct inode *inode, loff_t size);
Sage Weilde576062009-10-06 11:31:07 -0700663extern void __ceph_do_pending_vmtruncate(struct inode *inode);
Sage Weil3c6f6b72010-02-09 15:24:44 -0800664extern void ceph_queue_vmtruncate(struct inode *inode);
665
666extern void ceph_queue_invalidate(struct inode *inode);
667extern void ceph_queue_writeback(struct inode *inode);
Sage Weilde576062009-10-06 11:31:07 -0700668
669extern int ceph_do_getattr(struct inode *inode, int mask);
670extern int ceph_permission(struct inode *inode, int mask);
671extern int ceph_setattr(struct dentry *dentry, struct iattr *attr);
672extern int ceph_getattr(struct vfsmount *mnt, struct dentry *dentry,
673 struct kstat *stat);
674
675/* xattr.c */
676extern int ceph_setxattr(struct dentry *, const char *, const void *,
677 size_t, int);
678extern ssize_t ceph_getxattr(struct dentry *, const char *, void *, size_t);
679extern ssize_t ceph_listxattr(struct dentry *, char *, size_t);
680extern int ceph_removexattr(struct dentry *, const char *);
681extern void __ceph_build_xattrs_blob(struct ceph_inode_info *ci);
682extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci);
683
684/* caps.c */
685extern const char *ceph_cap_string(int c);
686extern void ceph_handle_caps(struct ceph_mds_session *session,
687 struct ceph_msg *msg);
688extern int ceph_add_cap(struct inode *inode,
689 struct ceph_mds_session *session, u64 cap_id,
690 int fmode, unsigned issued, unsigned wanted,
691 unsigned cap, unsigned seq, u64 realmino, int flags,
692 struct ceph_cap_reservation *caps_reservation);
Sage Weil7c1332b2010-02-16 11:39:45 -0800693extern void __ceph_remove_cap(struct ceph_cap *cap);
Sage Weilde576062009-10-06 11:31:07 -0700694static inline void ceph_remove_cap(struct ceph_cap *cap)
695{
696 struct inode *inode = &cap->ci->vfs_inode;
697 spin_lock(&inode->i_lock);
Sage Weil7c1332b2010-02-16 11:39:45 -0800698 __ceph_remove_cap(cap);
Sage Weilde576062009-10-06 11:31:07 -0700699 spin_unlock(&inode->i_lock);
700}
Yehuda Sadeh37151662010-06-17 16:16:12 -0700701extern void ceph_put_cap(struct ceph_mds_client *mdsc,
702 struct ceph_cap *cap);
Sage Weilde576062009-10-06 11:31:07 -0700703
704extern void ceph_queue_caps_release(struct inode *inode);
Stephen Rothwellf1a3d572010-01-18 11:53:08 +1100705extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc);
Christoph Hellwig7ea80852010-05-26 17:53:25 +0200706extern int ceph_fsync(struct file *file, int datasync);
Sage Weilde576062009-10-06 11:31:07 -0700707extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc,
708 struct ceph_mds_session *session);
Greg Farnum2bc50252010-06-30 12:44:34 -0700709extern struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci,
710 int mds);
Sage Weilde576062009-10-06 11:31:07 -0700711extern int ceph_get_cap_mds(struct inode *inode);
712extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps);
713extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had);
714extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
715 struct ceph_snap_context *snapc);
716extern void __ceph_flush_snaps(struct ceph_inode_info *ci,
Sage Weile8351242010-09-17 08:03:08 -0700717 struct ceph_mds_session **psession,
718 int again);
Sage Weilde576062009-10-06 11:31:07 -0700719extern void ceph_check_caps(struct ceph_inode_info *ci, int flags,
720 struct ceph_mds_session *session);
Sage Weilafcdaea2009-10-14 14:27:38 -0700721extern void ceph_check_delayed_caps(struct ceph_mds_client *mdsc);
722extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc);
Sage Weilde576062009-10-06 11:31:07 -0700723
724extern int ceph_encode_inode_release(void **p, struct inode *inode,
725 int mds, int drop, int unless, int force);
726extern int ceph_encode_dentry_release(void **p, struct dentry *dn,
727 int mds, int drop, int unless);
728
729extern int ceph_get_caps(struct ceph_inode_info *ci, int need, int want,
730 int *got, loff_t endoff);
731
732/* for counting open files by mode */
733static inline void __ceph_get_fmode(struct ceph_inode_info *ci, int mode)
734{
735 ci->i_nr_by_mode[mode]++;
736}
737extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode);
738
739/* addr.c */
740extern const struct address_space_operations ceph_aops;
741extern int ceph_mmap(struct file *file, struct vm_area_struct *vma);
742
743/* file.c */
744extern const struct file_operations ceph_file_fops;
745extern const struct address_space_operations ceph_aops;
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700746extern int ceph_copy_to_page_vector(struct page **pages,
747 const char *data,
748 loff_t off, size_t len);
749extern int ceph_copy_from_page_vector(struct page **pages,
750 char *data,
751 loff_t off, size_t len);
752extern struct page **ceph_alloc_page_vector(int num_pages, gfp_t flags);
Sage Weilde576062009-10-06 11:31:07 -0700753extern int ceph_open(struct inode *inode, struct file *file);
754extern struct dentry *ceph_lookup_open(struct inode *dir, struct dentry *dentry,
755 struct nameidata *nd, int mode,
756 int locked_dir);
757extern int ceph_release(struct inode *inode, struct file *filp);
Sage Weilde576062009-10-06 11:31:07 -0700758
759/* dir.c */
760extern const struct file_operations ceph_dir_fops;
761extern const struct inode_operations ceph_dir_iops;
Sage Weil52dfb8a2010-08-03 10:25:30 -0700762extern const struct dentry_operations ceph_dentry_ops, ceph_snap_dentry_ops,
Sage Weilde576062009-10-06 11:31:07 -0700763 ceph_snapdir_dentry_ops;
764
765extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry);
766extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req,
767 struct dentry *dentry, int err);
768
769extern void ceph_dentry_lru_add(struct dentry *dn);
770extern void ceph_dentry_lru_touch(struct dentry *dn);
771extern void ceph_dentry_lru_del(struct dentry *dn);
Sage Weil81a6cf22010-05-14 09:35:38 -0700772extern void ceph_invalidate_dentry_lease(struct dentry *dentry);
Sage Weilde576062009-10-06 11:31:07 -0700773
774/*
775 * our d_ops vary depending on whether the inode is live,
776 * snapshotted (read-only), or a virtual ".snap" directory.
777 */
778int ceph_init_dentry(struct dentry *dentry);
779
780
781/* ioctl.c */
782extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
783
784/* export.c */
785extern const struct export_operations ceph_export_ops;
786
Greg Farnum40819f62010-08-02 15:34:23 -0700787/* locks.c */
788extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl);
789extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl);
790extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num);
791extern int ceph_encode_locks(struct inode *i, struct ceph_pagelist *p,
792 int p_locks, int f_locks);
793extern int lock_to_ceph_filelock(struct file_lock *fl, struct ceph_filelock *c);
794
Sage Weilde576062009-10-06 11:31:07 -0700795static inline struct inode *get_dentry_parent_inode(struct dentry *dentry)
796{
797 if (dentry && dentry->d_parent)
798 return dentry->d_parent->d_inode;
799
800 return NULL;
801}
802
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700803/* debugfs.c */
804extern int ceph_fs_debugfs_init(struct ceph_fs_client *client);
805extern void ceph_fs_debugfs_cleanup(struct ceph_fs_client *client);
806
Sage Weilde576062009-10-06 11:31:07 -0700807#endif /* _FS_CEPH_SUPER_H */