blob: 5405c903704e8b4e481b18be9a7f173763283073 [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 */
Yehuda Sadehad1fee92011-01-21 16:44:03 -080030#define CEPH_MOUNT_OPT_INO32 (1<<8) /* 32 bit inos */
Sage Weil6a259382010-07-07 09:06:08 -070031
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070032#define CEPH_MOUNT_OPT_DEFAULT (CEPH_MOUNT_OPT_RBYTES)
Sage Weilde576062009-10-06 11:31:07 -070033
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070034#define ceph_set_mount_opt(fsc, opt) \
35 (fsc)->mount_options->flags |= CEPH_MOUNT_OPT_##opt;
36#define ceph_test_mount_opt(fsc, opt) \
37 (!!((fsc)->mount_options->flags & CEPH_MOUNT_OPT_##opt))
Sage Weilde576062009-10-06 11:31:07 -070038
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070039#define CEPH_MAX_READDIR_DEFAULT 1024
40#define CEPH_MAX_READDIR_BYTES_DEFAULT (512*1024)
41#define CEPH_SNAPDIRNAME_DEFAULT ".snap"
Sage Weilde576062009-10-06 11:31:07 -070042
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070043struct ceph_mount_options {
Sage Weilde576062009-10-06 11:31:07 -070044 int flags;
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070045 int sb_flags;
46
Sage Weil6e19a162010-04-29 16:38:32 -070047 int wsize;
48 int rsize; /* max readahead */
49 int congestion_kb; /* max writeback in flight */
50 int caps_wanted_delay_min, caps_wanted_delay_max;
51 int cap_release_safety;
Sage Weil23804d92010-05-14 13:06:30 -070052 int max_readdir; /* max readdir result (entires) */
53 int max_readdir_bytes; /* max readdir result (bytes) */
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070054
55 /*
56 * everything above this point can be memcmp'd; everything below
57 * is handled in compare_mount_options()
58 */
59
Sage Weilde576062009-10-06 11:31:07 -070060 char *snapdir_name; /* default ".snap" */
Sage Weilde576062009-10-06 11:31:07 -070061};
62
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070063struct ceph_fs_client {
Sage Weilde576062009-10-06 11:31:07 -070064 struct super_block *sb;
65
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070066 struct ceph_mount_options *mount_options;
67 struct ceph_client *client;
68
Sage Weilde576062009-10-06 11:31:07 -070069 unsigned long mount_state;
Sage Weil85ccce42010-02-17 10:02:43 -080070 int min_caps; /* min caps i added */
71
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070072 struct ceph_mds_client *mdsc;
Sage Weilde576062009-10-06 11:31:07 -070073
74 /* writeback */
75 mempool_t *wb_pagevec_pool;
76 struct workqueue_struct *wb_wq;
77 struct workqueue_struct *pg_inv_wq;
78 struct workqueue_struct *trunc_wq;
Yehuda Sadeh2baba252009-12-18 13:51:57 -080079 atomic_long_t writeback_count;
Sage Weilde576062009-10-06 11:31:07 -070080
81 struct backing_dev_info backing_dev_info;
Sage Weil07433042009-11-18 16:50:41 -080082
83#ifdef CONFIG_DEBUG_FS
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070084 struct dentry *debugfs_dentry_lru, *debugfs_caps;
Yehuda Sadeh2baba252009-12-18 13:51:57 -080085 struct dentry *debugfs_congestion_kb;
Sage Weil06edf042009-12-15 14:44:32 -080086 struct dentry *debugfs_bdi;
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070087 struct dentry *debugfs_mdsc, *debugfs_mdsmap;
Sage Weil07433042009-11-18 16:50:41 -080088#endif
Sage Weilde576062009-10-06 11:31:07 -070089};
90
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070091
Sage Weilde576062009-10-06 11:31:07 -070092/*
93 * File i/o capability. This tracks shared state with the metadata
94 * server that allows us to cache or writeback attributes or to read
95 * and write data. For any given inode, we should have one or more
96 * capabilities, one issued by each metadata server, and our
97 * cumulative access is the OR of all issued capabilities.
98 *
99 * Each cap is referenced by the inode's i_caps rbtree and by per-mds
100 * session capability lists.
101 */
102struct ceph_cap {
103 struct ceph_inode_info *ci;
104 struct rb_node ci_node; /* per-ci cap tree */
105 struct ceph_mds_session *session;
106 struct list_head session_caps; /* per-session caplist */
107 int mds;
108 u64 cap_id; /* unique cap id (mds provided) */
109 int issued; /* latest, from the mds */
110 int implemented; /* implemented superset of issued (for revocation) */
111 int mds_wanted;
Sage Weil685f9a5d2009-11-09 12:05:48 -0800112 u32 seq, issue_seq, mseq;
113 u32 cap_gen; /* active/stale cycle */
Sage Weilde576062009-10-06 11:31:07 -0700114 unsigned long last_used;
115 struct list_head caps_item;
116};
117
118#define CHECK_CAPS_NODELAY 1 /* do not delay any further */
119#define CHECK_CAPS_AUTHONLY 2 /* only check auth cap */
120#define CHECK_CAPS_FLUSH 4 /* flush any dirty caps */
121
122/*
123 * Snapped cap state that is pending flush to mds. When a snapshot occurs,
124 * we first complete any in-process sync writes and writeback any dirty
125 * data before flushing the snapped state (tracked here) back to the MDS.
126 */
127struct ceph_cap_snap {
128 atomic_t nref;
129 struct ceph_inode_info *ci;
130 struct list_head ci_item, flushing_item;
131
132 u64 follows, flush_tid;
133 int issued, dirty;
134 struct ceph_snap_context *context;
135
136 mode_t mode;
137 uid_t uid;
138 gid_t gid;
139
Sage Weil4a625be2010-08-22 15:03:56 -0700140 struct ceph_buffer *xattr_blob;
Sage Weilde576062009-10-06 11:31:07 -0700141 u64 xattr_version;
142
143 u64 size;
144 struct timespec mtime, atime, ctime;
145 u64 time_warp_seq;
146 int writing; /* a sync write is still in progress */
147 int dirty_pages; /* dirty pages awaiting writeback */
148};
149
150static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap)
151{
Sage Weil4a625be2010-08-22 15:03:56 -0700152 if (atomic_dec_and_test(&capsnap->nref)) {
153 if (capsnap->xattr_blob)
154 ceph_buffer_put(capsnap->xattr_blob);
Sage Weilde576062009-10-06 11:31:07 -0700155 kfree(capsnap);
Sage Weil4a625be2010-08-22 15:03:56 -0700156 }
Sage Weilde576062009-10-06 11:31:07 -0700157}
158
159/*
160 * The frag tree describes how a directory is fragmented, potentially across
161 * multiple metadata servers. It is also used to indicate points where
162 * metadata authority is delegated, and whether/where metadata is replicated.
163 *
164 * A _leaf_ frag will be present in the i_fragtree IFF there is
165 * delegation info. That is, if mds >= 0 || ndist > 0.
166 */
167#define CEPH_MAX_DIRFRAG_REP 4
168
169struct ceph_inode_frag {
170 struct rb_node node;
171
172 /* fragtree state */
173 u32 frag;
174 int split_by; /* i.e. 2^(split_by) children */
175
176 /* delegation and replication info */
177 int mds; /* -1 if same authority as parent */
178 int ndist; /* >0 if replicated */
179 int dist[CEPH_MAX_DIRFRAG_REP];
180};
181
182/*
183 * We cache inode xattrs as an encoded blob until they are first used,
184 * at which point we parse them into an rbtree.
185 */
186struct ceph_inode_xattr {
187 struct rb_node node;
188
189 const char *name;
190 int name_len;
191 const char *val;
192 int val_len;
193 int dirty;
194
195 int should_free_name;
196 int should_free_val;
197};
198
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700199/*
200 * Ceph dentry state
201 */
202struct ceph_dentry_info {
203 struct ceph_mds_session *lease_session;
204 u32 lease_gen, lease_shared_gen;
205 u32 lease_seq;
206 unsigned long lease_renew_after, lease_renew_from;
207 struct list_head lru;
208 struct dentry *dentry;
209 u64 time;
210 u64 offset;
211};
212
Sage Weilde576062009-10-06 11:31:07 -0700213struct ceph_inode_xattrs_info {
214 /*
215 * (still encoded) xattr blob. we avoid the overhead of parsing
216 * this until someone actually calls getxattr, etc.
217 *
218 * blob->vec.iov_len == 4 implies there are no xattrs; blob ==
219 * NULL means we don't know.
220 */
221 struct ceph_buffer *blob, *prealloc_blob;
222
223 struct rb_root index;
224 bool dirty;
225 int count;
226 int names_size;
227 int vals_size;
228 u64 version, index_version;
229};
230
231/*
232 * Ceph inode.
233 */
Sage Weilde576062009-10-06 11:31:07 -0700234struct ceph_inode_info {
235 struct ceph_vino i_vino; /* ceph ino + snap */
236
237 u64 i_version;
238 u32 i_time_warp_seq;
239
240 unsigned i_ceph_flags;
241 unsigned long i_release_count;
242
Sage Weil6c0f3af2010-11-16 11:14:34 -0800243 struct ceph_dir_layout i_dir_layout;
Sage Weilde576062009-10-06 11:31:07 -0700244 struct ceph_file_layout i_layout;
245 char *i_symlink;
246
247 /* for dirs */
248 struct timespec i_rctime;
249 u64 i_rbytes, i_rfiles, i_rsubdirs;
250 u64 i_files, i_subdirs;
251 u64 i_max_offset; /* largest readdir offset, set with I_COMPLETE */
252
253 struct rb_root i_fragtree;
254 struct mutex i_fragtree_mutex;
255
256 struct ceph_inode_xattrs_info i_xattrs;
257
258 /* capabilities. protected _both_ by i_lock and cap->session's
259 * s_mutex. */
260 struct rb_root i_caps; /* cap list */
261 struct ceph_cap *i_auth_cap; /* authoritative cap, if any */
262 unsigned i_dirty_caps, i_flushing_caps; /* mask of dirtied fields */
263 struct list_head i_dirty_item, i_flushing_item;
264 u64 i_cap_flush_seq;
265 /* we need to track cap writeback on a per-cap-bit basis, to allow
266 * overlapping, pipelined cap flushes to the mds. we can probably
267 * reduce the tid to 8 bits if we're concerned about inode size. */
268 u16 i_cap_flush_last_tid, i_cap_flush_tid[CEPH_CAP_BITS];
269 wait_queue_head_t i_cap_wq; /* threads waiting on a capability */
270 unsigned long i_hold_caps_min; /* jiffies */
271 unsigned long i_hold_caps_max; /* jiffies */
272 struct list_head i_cap_delay_list; /* for delayed cap release to mds */
273 int i_cap_exporting_mds; /* to handle cap migration between */
274 unsigned i_cap_exporting_mseq; /* mds's. */
275 unsigned i_cap_exporting_issued;
276 struct ceph_cap_reservation i_cap_migration_resv;
277 struct list_head i_cap_snaps; /* snapped state pending flush to mds */
Sage Weil7d8cb262010-08-24 08:44:16 -0700278 struct ceph_snap_context *i_head_snapc; /* set if wr_buffer_head > 0 or
279 dirty|flushing caps */
Sage Weilde576062009-10-06 11:31:07 -0700280 unsigned i_snap_caps; /* cap bits for snapped files */
281
282 int i_nr_by_mode[CEPH_FILE_MODE_NUM]; /* open file counts */
283
284 u32 i_truncate_seq; /* last truncate to smaller size */
285 u64 i_truncate_size; /* and the size we last truncated down to */
286 int i_truncate_pending; /* still need to call vmtruncate */
287
288 u64 i_max_size; /* max file size authorized by mds */
289 u64 i_reported_size; /* (max_)size reported to or requested of mds */
290 u64 i_wanted_max_size; /* offset we'd like to write too */
291 u64 i_requested_max_size; /* max_size we've requested */
292
293 /* held references to caps */
294 int i_pin_ref;
295 int i_rd_ref, i_rdcache_ref, i_wr_ref;
296 int i_wrbuffer_ref, i_wrbuffer_ref_head;
297 u32 i_shared_gen; /* increment each time we get FILE_SHARED */
Sage Weilcd045cb2010-11-04 11:05:05 -0700298 u32 i_rdcache_gen; /* incremented each time we get FILE_CACHE. */
Sage Weilde576062009-10-06 11:31:07 -0700299 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 Sadehad1fee92011-01-21 16:44:03 -0800323static inline struct ceph_fs_client *ceph_inode_to_client(struct inode *inode)
324{
325 return (struct ceph_fs_client *)inode->i_sb->s_fs_info;
326}
327
328static inline struct ceph_fs_client *ceph_sb_to_client(struct super_block *sb)
329{
330 return (struct ceph_fs_client *)sb->s_fs_info;
331}
332
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700333static inline struct ceph_vino ceph_vino(struct inode *inode)
334{
335 return ceph_inode(inode)->i_vino;
336}
337
338/*
339 * ino_t is <64 bits on many architectures, blech.
340 *
Yehuda Sadehad1fee92011-01-21 16:44:03 -0800341 * i_ino (kernel inode) st_ino (userspace)
342 * i386 32 32
343 * x86_64+ino32 64 32
344 * x86_64 64 64
345 */
346static inline u32 ceph_ino_to_ino32(ino_t ino)
347{
348 ino ^= ino >> (sizeof(ino) * 8 - 32);
349 if (!ino)
350 ino = 1;
351 return ino;
352}
353
354/*
355 * kernel i_ino value
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700356 */
357static inline ino_t ceph_vino_to_ino(struct ceph_vino vino)
358{
359 ino_t ino = (ino_t)vino.ino; /* ^ (vino.snap << 20); */
360#if BITS_PER_LONG == 32
Yehuda Sadehad1fee92011-01-21 16:44:03 -0800361 ino = ceph_ino_to_ino32(ino);
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700362#endif
363 return ino;
364}
365
Yehuda Sadehad1fee92011-01-21 16:44:03 -0800366/*
367 * user-visible ino (stat, filldir)
368 */
369#if BITS_PER_LONG == 32
370static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino)
371{
372 return ino;
373}
374#else
375static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino)
376{
377 if (ceph_test_mount_opt(ceph_sb_to_client(sb), INO32))
378 ino = ceph_ino_to_ino32(ino);
379 return ino;
380}
381#endif
382
383
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700384/* for printf-style formatting */
385#define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap
386
387static inline u64 ceph_ino(struct inode *inode)
388{
389 return ceph_inode(inode)->i_vino.ino;
390}
391static inline u64 ceph_snap(struct inode *inode)
392{
393 return ceph_inode(inode)->i_vino.snap;
394}
395
396static inline int ceph_ino_compare(struct inode *inode, void *data)
397{
398 struct ceph_vino *pvino = (struct ceph_vino *)data;
399 struct ceph_inode_info *ci = ceph_inode(inode);
400 return ci->i_vino.ino == pvino->ino &&
401 ci->i_vino.snap == pvino->snap;
402}
403
404static inline struct inode *ceph_find_inode(struct super_block *sb,
405 struct ceph_vino vino)
406{
407 ino_t t = ceph_vino_to_ino(vino);
408 return ilookup5(sb, t, ceph_ino_compare, &vino);
409}
410
411
412/*
413 * Ceph inode.
414 */
415#define CEPH_I_COMPLETE 1 /* we have complete directory cached */
416#define CEPH_I_NODELAY 4 /* do not delay cap release */
417#define CEPH_I_FLUSH 8 /* do not delay flush of dirty metadata */
418#define CEPH_I_NOFLUSH 16 /* do not flush dirty caps */
419
Sage Weilde576062009-10-06 11:31:07 -0700420static inline void ceph_i_clear(struct inode *inode, unsigned mask)
421{
422 struct ceph_inode_info *ci = ceph_inode(inode);
423
424 spin_lock(&inode->i_lock);
425 ci->i_ceph_flags &= ~mask;
426 spin_unlock(&inode->i_lock);
427}
428
429static inline void ceph_i_set(struct inode *inode, unsigned mask)
430{
431 struct ceph_inode_info *ci = ceph_inode(inode);
432
433 spin_lock(&inode->i_lock);
434 ci->i_ceph_flags |= mask;
435 spin_unlock(&inode->i_lock);
436}
437
438static inline bool ceph_i_test(struct inode *inode, unsigned mask)
439{
440 struct ceph_inode_info *ci = ceph_inode(inode);
441 bool r;
442
Sage Weilefa4c122010-10-18 14:04:31 -0700443 spin_lock(&inode->i_lock);
Sage Weilde576062009-10-06 11:31:07 -0700444 r = (ci->i_ceph_flags & mask) == mask;
Sage Weilefa4c122010-10-18 14:04:31 -0700445 spin_unlock(&inode->i_lock);
Sage Weilde576062009-10-06 11:31:07 -0700446 return r;
447}
448
449
450/* find a specific frag @f */
451extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci,
452 u32 f);
453
454/*
455 * choose fragment for value @v. copy frag content to pfrag, if leaf
456 * exists
457 */
458extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v,
459 struct ceph_inode_frag *pfrag,
460 int *found);
461
Sage Weilde576062009-10-06 11:31:07 -0700462static inline struct ceph_dentry_info *ceph_dentry(struct dentry *dentry)
463{
464 return (struct ceph_dentry_info *)dentry->d_fsdata;
465}
466
467static inline loff_t ceph_make_fpos(unsigned frag, unsigned off)
468{
469 return ((loff_t)frag << 32) | (loff_t)off;
470}
471
Sage Weilde576062009-10-06 11:31:07 -0700472/*
473 * caps helpers
474 */
475static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci)
476{
477 return !RB_EMPTY_ROOT(&ci->i_caps);
478}
479
480extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented);
481extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t);
482extern int __ceph_caps_issued_other(struct ceph_inode_info *ci,
483 struct ceph_cap *cap);
484
485static inline int ceph_caps_issued(struct ceph_inode_info *ci)
486{
487 int issued;
488 spin_lock(&ci->vfs_inode.i_lock);
489 issued = __ceph_caps_issued(ci, NULL);
490 spin_unlock(&ci->vfs_inode.i_lock);
491 return issued;
492}
493
494static inline int ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask,
495 int touch)
496{
497 int r;
498 spin_lock(&ci->vfs_inode.i_lock);
499 r = __ceph_caps_issued_mask(ci, mask, touch);
500 spin_unlock(&ci->vfs_inode.i_lock);
501 return r;
502}
503
504static inline int __ceph_caps_dirty(struct ceph_inode_info *ci)
505{
506 return ci->i_dirty_caps | ci->i_flushing_caps;
507}
Sage Weilafcdaea2009-10-14 14:27:38 -0700508extern void __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask);
Sage Weilde576062009-10-06 11:31:07 -0700509
510extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask);
511extern int __ceph_caps_used(struct ceph_inode_info *ci);
512
513extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci);
514
515/*
516 * wanted, by virtue of open file modes AND cap refs (buffered/cached data)
517 */
518static inline int __ceph_caps_wanted(struct ceph_inode_info *ci)
519{
520 int w = __ceph_caps_file_wanted(ci) | __ceph_caps_used(ci);
521 if (w & CEPH_CAP_FILE_BUFFER)
522 w |= CEPH_CAP_FILE_EXCL; /* we want EXCL if dirty data */
523 return w;
524}
525
526/* what the mds thinks we want */
527extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci);
528
Yehuda Sadeh37151662010-06-17 16:16:12 -0700529extern void ceph_caps_init(struct ceph_mds_client *mdsc);
530extern void ceph_caps_finalize(struct ceph_mds_client *mdsc);
531extern void ceph_adjust_min_caps(struct ceph_mds_client *mdsc, int delta);
532extern int ceph_reserve_caps(struct ceph_mds_client *mdsc,
533 struct ceph_cap_reservation *ctx, int need);
534extern int ceph_unreserve_caps(struct ceph_mds_client *mdsc,
535 struct ceph_cap_reservation *ctx);
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700536extern void ceph_reservation_status(struct ceph_fs_client *client,
Sage Weilde576062009-10-06 11:31:07 -0700537 int *total, int *avail, int *used,
Sage Weil85ccce42010-02-17 10:02:43 -0800538 int *reserved, int *min);
Sage Weilde576062009-10-06 11:31:07 -0700539
Sage Weilde576062009-10-06 11:31:07 -0700540
Sage Weilde576062009-10-06 11:31:07 -0700541
542/*
543 * we keep buffered readdir results attached to file->private_data
544 */
545struct ceph_file_info {
546 int fmode; /* initialized on open */
547
548 /* readdir: position within the dir */
549 u32 frag;
550 struct ceph_mds_request *last_readdir;
551 int at_end;
552
553 /* readdir: position within a frag */
554 unsigned offset; /* offset of last chunk, adjusted for . and .. */
555 u64 next_offset; /* offset of next chunk (last_name's + 1) */
556 char *last_name; /* last entry in previous chunk */
557 struct dentry *dentry; /* next dentry (for dcache readdir) */
558 unsigned long dir_release_count;
559
560 /* used for -o dirstat read() on directory thing */
561 char *dir_info;
562 int dir_info_len;
563};
564
565
566
567/*
Sage Weilde576062009-10-06 11:31:07 -0700568 * A "snap realm" describes a subset of the file hierarchy sharing
569 * the same set of snapshots that apply to it. The realms themselves
570 * are organized into a hierarchy, such that children inherit (some of)
571 * the snapshots of their parents.
572 *
573 * All inodes within the realm that have capabilities are linked into a
574 * per-realm list.
575 */
576struct ceph_snap_realm {
577 u64 ino;
578 atomic_t nref;
Sage Weila105f002010-02-15 14:37:55 -0800579 struct rb_node node;
580
Sage Weilde576062009-10-06 11:31:07 -0700581 u64 created, seq;
582 u64 parent_ino;
583 u64 parent_since; /* snapid when our current parent became so */
584
585 u64 *prior_parent_snaps; /* snaps inherited from any parents we */
586 int num_prior_parent_snaps; /* had prior to parent_since */
587 u64 *snaps; /* snaps specific to this realm */
588 int num_snaps;
589
590 struct ceph_snap_realm *parent;
591 struct list_head children; /* list of child realms */
592 struct list_head child_item;
593
594 struct list_head empty_item; /* if i have ref==0 */
595
Sage Weilae00d4f2010-09-16 16:26:51 -0700596 struct list_head dirty_item; /* if realm needs new context */
597
Sage Weilde576062009-10-06 11:31:07 -0700598 /* the current set of snaps for this realm */
599 struct ceph_snap_context *cached_context;
600
601 struct list_head inodes_with_caps;
602 spinlock_t inodes_with_caps_lock;
603};
604
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700605static inline int default_congestion_kb(void)
Sage Weilde576062009-10-06 11:31:07 -0700606{
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700607 int congestion_kb;
608
609 /*
610 * Copied from NFS
611 *
612 * congestion size, scale with available memory.
613 *
614 * 64MB: 8192k
615 * 128MB: 11585k
616 * 256MB: 16384k
617 * 512MB: 23170k
618 * 1GB: 32768k
619 * 2GB: 46340k
620 * 4GB: 65536k
621 * 8GB: 92681k
622 * 16GB: 131072k
623 *
624 * This allows larger machines to have larger/more transfers.
625 * Limit the default to 256M
626 */
627 congestion_kb = (16*int_sqrt(totalram_pages)) << (PAGE_SHIFT-10);
628 if (congestion_kb > 256*1024)
629 congestion_kb = 256*1024;
630
631 return congestion_kb;
Sage Weilde576062009-10-06 11:31:07 -0700632}
633
634
635
636/* snap.c */
637struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc,
638 u64 ino);
639extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc,
640 struct ceph_snap_realm *realm);
641extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc,
642 struct ceph_snap_realm *realm);
643extern int ceph_update_snap_trace(struct ceph_mds_client *m,
644 void *p, void *e, bool deletion);
645extern void ceph_handle_snap(struct ceph_mds_client *mdsc,
Sage Weil2600d2d2010-02-22 15:12:16 -0800646 struct ceph_mds_session *session,
Sage Weilde576062009-10-06 11:31:07 -0700647 struct ceph_msg *msg);
Sage Weilfc837c8f2010-04-13 11:41:22 -0700648extern void ceph_queue_cap_snap(struct ceph_inode_info *ci);
Sage Weilde576062009-10-06 11:31:07 -0700649extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci,
650 struct ceph_cap_snap *capsnap);
651extern void ceph_cleanup_empty_realms(struct ceph_mds_client *mdsc);
652
653/*
654 * a cap_snap is "pending" if it is still awaiting an in-progress
655 * sync write (that may/may not still update size, mtime, etc.).
656 */
657static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci)
658{
659 return !list_empty(&ci->i_cap_snaps) &&
660 list_entry(ci->i_cap_snaps.prev, struct ceph_cap_snap,
661 ci_item)->writing;
662}
663
Sage Weilde576062009-10-06 11:31:07 -0700664/* inode.c */
665extern const struct inode_operations ceph_file_iops;
666
667extern struct inode *ceph_alloc_inode(struct super_block *sb);
668extern void ceph_destroy_inode(struct inode *inode);
669
670extern struct inode *ceph_get_inode(struct super_block *sb,
671 struct ceph_vino vino);
672extern struct inode *ceph_get_snapdir(struct inode *parent);
673extern int ceph_fill_file_size(struct inode *inode, int issued,
674 u32 truncate_seq, u64 truncate_size, u64 size);
675extern void ceph_fill_file_time(struct inode *inode, int issued,
676 u64 time_warp_seq, struct timespec *ctime,
677 struct timespec *mtime, struct timespec *atime);
678extern int ceph_fill_trace(struct super_block *sb,
679 struct ceph_mds_request *req,
680 struct ceph_mds_session *session);
681extern int ceph_readdir_prepopulate(struct ceph_mds_request *req,
682 struct ceph_mds_session *session);
683
684extern int ceph_inode_holds_cap(struct inode *inode, int mask);
685
686extern int ceph_inode_set_size(struct inode *inode, loff_t size);
Sage Weilde576062009-10-06 11:31:07 -0700687extern void __ceph_do_pending_vmtruncate(struct inode *inode);
Sage Weil3c6f6b72010-02-09 15:24:44 -0800688extern void ceph_queue_vmtruncate(struct inode *inode);
689
690extern void ceph_queue_invalidate(struct inode *inode);
691extern void ceph_queue_writeback(struct inode *inode);
Sage Weilde576062009-10-06 11:31:07 -0700692
693extern int ceph_do_getattr(struct inode *inode, int mask);
Nick Pigginb74c79e2011-01-07 17:49:58 +1100694extern int ceph_permission(struct inode *inode, int mask, unsigned int flags);
Sage Weilde576062009-10-06 11:31:07 -0700695extern int ceph_setattr(struct dentry *dentry, struct iattr *attr);
696extern int ceph_getattr(struct vfsmount *mnt, struct dentry *dentry,
697 struct kstat *stat);
698
699/* xattr.c */
700extern int ceph_setxattr(struct dentry *, const char *, const void *,
701 size_t, int);
702extern ssize_t ceph_getxattr(struct dentry *, const char *, void *, size_t);
703extern ssize_t ceph_listxattr(struct dentry *, char *, size_t);
704extern int ceph_removexattr(struct dentry *, const char *);
705extern void __ceph_build_xattrs_blob(struct ceph_inode_info *ci);
706extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci);
707
708/* caps.c */
709extern const char *ceph_cap_string(int c);
710extern void ceph_handle_caps(struct ceph_mds_session *session,
711 struct ceph_msg *msg);
712extern int ceph_add_cap(struct inode *inode,
713 struct ceph_mds_session *session, u64 cap_id,
714 int fmode, unsigned issued, unsigned wanted,
715 unsigned cap, unsigned seq, u64 realmino, int flags,
716 struct ceph_cap_reservation *caps_reservation);
Sage Weil7c1332b2010-02-16 11:39:45 -0800717extern void __ceph_remove_cap(struct ceph_cap *cap);
Sage Weilde576062009-10-06 11:31:07 -0700718static inline void ceph_remove_cap(struct ceph_cap *cap)
719{
720 struct inode *inode = &cap->ci->vfs_inode;
721 spin_lock(&inode->i_lock);
Sage Weil7c1332b2010-02-16 11:39:45 -0800722 __ceph_remove_cap(cap);
Sage Weilde576062009-10-06 11:31:07 -0700723 spin_unlock(&inode->i_lock);
724}
Yehuda Sadeh37151662010-06-17 16:16:12 -0700725extern void ceph_put_cap(struct ceph_mds_client *mdsc,
726 struct ceph_cap *cap);
Sage Weilde576062009-10-06 11:31:07 -0700727
728extern void ceph_queue_caps_release(struct inode *inode);
Stephen Rothwellf1a3d572010-01-18 11:53:08 +1100729extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc);
Christoph Hellwig7ea80852010-05-26 17:53:25 +0200730extern int ceph_fsync(struct file *file, int datasync);
Sage Weilde576062009-10-06 11:31:07 -0700731extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc,
732 struct ceph_mds_session *session);
Greg Farnum2bc50252010-06-30 12:44:34 -0700733extern struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci,
734 int mds);
Sage Weilde576062009-10-06 11:31:07 -0700735extern int ceph_get_cap_mds(struct inode *inode);
736extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps);
737extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had);
738extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
739 struct ceph_snap_context *snapc);
740extern void __ceph_flush_snaps(struct ceph_inode_info *ci,
Sage Weile8351242010-09-17 08:03:08 -0700741 struct ceph_mds_session **psession,
742 int again);
Sage Weilde576062009-10-06 11:31:07 -0700743extern void ceph_check_caps(struct ceph_inode_info *ci, int flags,
744 struct ceph_mds_session *session);
Sage Weilafcdaea2009-10-14 14:27:38 -0700745extern void ceph_check_delayed_caps(struct ceph_mds_client *mdsc);
746extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc);
Sage Weilde576062009-10-06 11:31:07 -0700747
748extern int ceph_encode_inode_release(void **p, struct inode *inode,
749 int mds, int drop, int unless, int force);
750extern int ceph_encode_dentry_release(void **p, struct dentry *dn,
751 int mds, int drop, int unless);
752
753extern int ceph_get_caps(struct ceph_inode_info *ci, int need, int want,
754 int *got, loff_t endoff);
755
756/* for counting open files by mode */
757static inline void __ceph_get_fmode(struct ceph_inode_info *ci, int mode)
758{
759 ci->i_nr_by_mode[mode]++;
760}
761extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode);
762
763/* addr.c */
764extern const struct address_space_operations ceph_aops;
765extern int ceph_mmap(struct file *file, struct vm_area_struct *vma);
766
767/* file.c */
768extern const struct file_operations ceph_file_fops;
769extern const struct address_space_operations ceph_aops;
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700770extern int ceph_copy_to_page_vector(struct page **pages,
771 const char *data,
772 loff_t off, size_t len);
773extern int ceph_copy_from_page_vector(struct page **pages,
774 char *data,
775 loff_t off, size_t len);
776extern struct page **ceph_alloc_page_vector(int num_pages, gfp_t flags);
Sage Weilde576062009-10-06 11:31:07 -0700777extern int ceph_open(struct inode *inode, struct file *file);
778extern struct dentry *ceph_lookup_open(struct inode *dir, struct dentry *dentry,
779 struct nameidata *nd, int mode,
780 int locked_dir);
781extern int ceph_release(struct inode *inode, struct file *filp);
Sage Weilde576062009-10-06 11:31:07 -0700782
783/* dir.c */
784extern const struct file_operations ceph_dir_fops;
785extern const struct inode_operations ceph_dir_iops;
Sage Weil52dfb8a2010-08-03 10:25:30 -0700786extern const struct dentry_operations ceph_dentry_ops, ceph_snap_dentry_ops,
Sage Weilde576062009-10-06 11:31:07 -0700787 ceph_snapdir_dentry_ops;
788
789extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry);
790extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req,
791 struct dentry *dentry, int err);
792
793extern void ceph_dentry_lru_add(struct dentry *dn);
794extern void ceph_dentry_lru_touch(struct dentry *dn);
795extern void ceph_dentry_lru_del(struct dentry *dn);
Sage Weil81a6cf22010-05-14 09:35:38 -0700796extern void ceph_invalidate_dentry_lease(struct dentry *dentry);
Sage Weil6c0f3af2010-11-16 11:14:34 -0800797extern unsigned ceph_dentry_hash(struct dentry *dn);
Sage Weilde576062009-10-06 11:31:07 -0700798
799/*
800 * our d_ops vary depending on whether the inode is live,
801 * snapshotted (read-only), or a virtual ".snap" directory.
802 */
803int ceph_init_dentry(struct dentry *dentry);
804
805
806/* ioctl.c */
807extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
808
809/* export.c */
810extern const struct export_operations ceph_export_ops;
811
Greg Farnum40819f62010-08-02 15:34:23 -0700812/* locks.c */
813extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl);
814extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl);
815extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num);
816extern int ceph_encode_locks(struct inode *i, struct ceph_pagelist *p,
817 int p_locks, int f_locks);
818extern int lock_to_ceph_filelock(struct file_lock *fl, struct ceph_filelock *c);
819
Sage Weilde576062009-10-06 11:31:07 -0700820static inline struct inode *get_dentry_parent_inode(struct dentry *dentry)
821{
822 if (dentry && dentry->d_parent)
823 return dentry->d_parent->d_inode;
824
825 return NULL;
826}
827
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700828/* debugfs.c */
829extern int ceph_fs_debugfs_init(struct ceph_fs_client *client);
830extern void ceph_fs_debugfs_cleanup(struct ceph_fs_client *client);
831
Sage Weilde576062009-10-06 11:31:07 -0700832#endif /* _FS_CEPH_SUPER_H */