blob: cdb17d36755c4a67ade569afb16a724031cc2530 [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
Sage Weil80456f82011-03-10 13:33:26 -080039#define CEPH_RSIZE_DEFAULT (512*1024) /* readahead */
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070040#define CEPH_MAX_READDIR_DEFAULT 1024
41#define CEPH_MAX_READDIR_BYTES_DEFAULT (512*1024)
42#define CEPH_SNAPDIRNAME_DEFAULT ".snap"
Sage Weilde576062009-10-06 11:31:07 -070043
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070044struct ceph_mount_options {
Sage Weilde576062009-10-06 11:31:07 -070045 int flags;
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070046 int sb_flags;
47
Sage Weil6e19a162010-04-29 16:38:32 -070048 int wsize;
49 int rsize; /* max readahead */
50 int congestion_kb; /* max writeback in flight */
51 int caps_wanted_delay_min, caps_wanted_delay_max;
52 int cap_release_safety;
Sage Weil23804d92010-05-14 13:06:30 -070053 int max_readdir; /* max readdir result (entires) */
54 int max_readdir_bytes; /* max readdir result (bytes) */
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070055
56 /*
57 * everything above this point can be memcmp'd; everything below
58 * is handled in compare_mount_options()
59 */
60
Sage Weilde576062009-10-06 11:31:07 -070061 char *snapdir_name; /* default ".snap" */
Sage Weilde576062009-10-06 11:31:07 -070062};
63
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070064struct ceph_fs_client {
Sage Weilde576062009-10-06 11:31:07 -070065 struct super_block *sb;
66
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070067 struct ceph_mount_options *mount_options;
68 struct ceph_client *client;
69
Sage Weilde576062009-10-06 11:31:07 -070070 unsigned long mount_state;
Sage Weil85ccce42010-02-17 10:02:43 -080071 int min_caps; /* min caps i added */
72
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070073 struct ceph_mds_client *mdsc;
Sage Weilde576062009-10-06 11:31:07 -070074
75 /* writeback */
76 mempool_t *wb_pagevec_pool;
77 struct workqueue_struct *wb_wq;
78 struct workqueue_struct *pg_inv_wq;
79 struct workqueue_struct *trunc_wq;
Yehuda Sadeh2baba252009-12-18 13:51:57 -080080 atomic_long_t writeback_count;
Sage Weilde576062009-10-06 11:31:07 -070081
82 struct backing_dev_info backing_dev_info;
Sage Weil07433042009-11-18 16:50:41 -080083
84#ifdef CONFIG_DEBUG_FS
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070085 struct dentry *debugfs_dentry_lru, *debugfs_caps;
Yehuda Sadeh2baba252009-12-18 13:51:57 -080086 struct dentry *debugfs_congestion_kb;
Sage Weil06edf042009-12-15 14:44:32 -080087 struct dentry *debugfs_bdi;
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070088 struct dentry *debugfs_mdsc, *debugfs_mdsmap;
Sage Weil07433042009-11-18 16:50:41 -080089#endif
Sage Weilde576062009-10-06 11:31:07 -070090};
91
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070092
Sage Weilde576062009-10-06 11:31:07 -070093/*
94 * File i/o capability. This tracks shared state with the metadata
95 * server that allows us to cache or writeback attributes or to read
96 * and write data. For any given inode, we should have one or more
97 * capabilities, one issued by each metadata server, and our
98 * cumulative access is the OR of all issued capabilities.
99 *
100 * Each cap is referenced by the inode's i_caps rbtree and by per-mds
101 * session capability lists.
102 */
103struct ceph_cap {
104 struct ceph_inode_info *ci;
105 struct rb_node ci_node; /* per-ci cap tree */
106 struct ceph_mds_session *session;
107 struct list_head session_caps; /* per-session caplist */
108 int mds;
109 u64 cap_id; /* unique cap id (mds provided) */
110 int issued; /* latest, from the mds */
111 int implemented; /* implemented superset of issued (for revocation) */
112 int mds_wanted;
Sage Weil685f9a5d2009-11-09 12:05:48 -0800113 u32 seq, issue_seq, mseq;
114 u32 cap_gen; /* active/stale cycle */
Sage Weilde576062009-10-06 11:31:07 -0700115 unsigned long last_used;
116 struct list_head caps_item;
117};
118
119#define CHECK_CAPS_NODELAY 1 /* do not delay any further */
120#define CHECK_CAPS_AUTHONLY 2 /* only check auth cap */
121#define CHECK_CAPS_FLUSH 4 /* flush any dirty caps */
122
123/*
124 * Snapped cap state that is pending flush to mds. When a snapshot occurs,
125 * we first complete any in-process sync writes and writeback any dirty
126 * data before flushing the snapped state (tracked here) back to the MDS.
127 */
128struct ceph_cap_snap {
129 atomic_t nref;
130 struct ceph_inode_info *ci;
131 struct list_head ci_item, flushing_item;
132
133 u64 follows, flush_tid;
134 int issued, dirty;
135 struct ceph_snap_context *context;
136
137 mode_t mode;
138 uid_t uid;
139 gid_t gid;
140
Sage Weil4a625be2010-08-22 15:03:56 -0700141 struct ceph_buffer *xattr_blob;
Sage Weilde576062009-10-06 11:31:07 -0700142 u64 xattr_version;
143
144 u64 size;
145 struct timespec mtime, atime, ctime;
146 u64 time_warp_seq;
147 int writing; /* a sync write is still in progress */
148 int dirty_pages; /* dirty pages awaiting writeback */
149};
150
151static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap)
152{
Sage Weil4a625be2010-08-22 15:03:56 -0700153 if (atomic_dec_and_test(&capsnap->nref)) {
154 if (capsnap->xattr_blob)
155 ceph_buffer_put(capsnap->xattr_blob);
Sage Weilde576062009-10-06 11:31:07 -0700156 kfree(capsnap);
Sage Weil4a625be2010-08-22 15:03:56 -0700157 }
Sage Weilde576062009-10-06 11:31:07 -0700158}
159
160/*
161 * The frag tree describes how a directory is fragmented, potentially across
162 * multiple metadata servers. It is also used to indicate points where
163 * metadata authority is delegated, and whether/where metadata is replicated.
164 *
165 * A _leaf_ frag will be present in the i_fragtree IFF there is
166 * delegation info. That is, if mds >= 0 || ndist > 0.
167 */
168#define CEPH_MAX_DIRFRAG_REP 4
169
170struct ceph_inode_frag {
171 struct rb_node node;
172
173 /* fragtree state */
174 u32 frag;
175 int split_by; /* i.e. 2^(split_by) children */
176
177 /* delegation and replication info */
178 int mds; /* -1 if same authority as parent */
179 int ndist; /* >0 if replicated */
180 int dist[CEPH_MAX_DIRFRAG_REP];
181};
182
183/*
184 * We cache inode xattrs as an encoded blob until they are first used,
185 * at which point we parse them into an rbtree.
186 */
187struct ceph_inode_xattr {
188 struct rb_node node;
189
190 const char *name;
191 int name_len;
192 const char *val;
193 int val_len;
194 int dirty;
195
196 int should_free_name;
197 int should_free_val;
198};
199
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700200/*
201 * Ceph dentry state
202 */
203struct ceph_dentry_info {
204 struct ceph_mds_session *lease_session;
205 u32 lease_gen, lease_shared_gen;
206 u32 lease_seq;
207 unsigned long lease_renew_after, lease_renew_from;
208 struct list_head lru;
209 struct dentry *dentry;
210 u64 time;
211 u64 offset;
212};
213
Sage Weilde576062009-10-06 11:31:07 -0700214struct ceph_inode_xattrs_info {
215 /*
216 * (still encoded) xattr blob. we avoid the overhead of parsing
217 * this until someone actually calls getxattr, etc.
218 *
219 * blob->vec.iov_len == 4 implies there are no xattrs; blob ==
220 * NULL means we don't know.
221 */
222 struct ceph_buffer *blob, *prealloc_blob;
223
224 struct rb_root index;
225 bool dirty;
226 int count;
227 int names_size;
228 int vals_size;
229 u64 version, index_version;
230};
231
232/*
233 * Ceph inode.
234 */
Sage Weilde576062009-10-06 11:31:07 -0700235struct ceph_inode_info {
236 struct ceph_vino i_vino; /* ceph ino + snap */
237
238 u64 i_version;
239 u32 i_time_warp_seq;
240
241 unsigned i_ceph_flags;
242 unsigned long i_release_count;
243
Sage Weil6c0f3af2010-11-16 11:14:34 -0800244 struct ceph_dir_layout i_dir_layout;
Sage Weilde576062009-10-06 11:31:07 -0700245 struct ceph_file_layout i_layout;
246 char *i_symlink;
247
248 /* for dirs */
249 struct timespec i_rctime;
250 u64 i_rbytes, i_rfiles, i_rsubdirs;
251 u64 i_files, i_subdirs;
252 u64 i_max_offset; /* largest readdir offset, set with I_COMPLETE */
253
254 struct rb_root i_fragtree;
255 struct mutex i_fragtree_mutex;
256
257 struct ceph_inode_xattrs_info i_xattrs;
258
259 /* capabilities. protected _both_ by i_lock and cap->session's
260 * s_mutex. */
261 struct rb_root i_caps; /* cap list */
262 struct ceph_cap *i_auth_cap; /* authoritative cap, if any */
263 unsigned i_dirty_caps, i_flushing_caps; /* mask of dirtied fields */
264 struct list_head i_dirty_item, i_flushing_item;
265 u64 i_cap_flush_seq;
266 /* we need to track cap writeback on a per-cap-bit basis, to allow
267 * overlapping, pipelined cap flushes to the mds. we can probably
268 * reduce the tid to 8 bits if we're concerned about inode size. */
269 u16 i_cap_flush_last_tid, i_cap_flush_tid[CEPH_CAP_BITS];
270 wait_queue_head_t i_cap_wq; /* threads waiting on a capability */
271 unsigned long i_hold_caps_min; /* jiffies */
272 unsigned long i_hold_caps_max; /* jiffies */
273 struct list_head i_cap_delay_list; /* for delayed cap release to mds */
274 int i_cap_exporting_mds; /* to handle cap migration between */
275 unsigned i_cap_exporting_mseq; /* mds's. */
276 unsigned i_cap_exporting_issued;
277 struct ceph_cap_reservation i_cap_migration_resv;
278 struct list_head i_cap_snaps; /* snapped state pending flush to mds */
Sage Weil7d8cb262010-08-24 08:44:16 -0700279 struct ceph_snap_context *i_head_snapc; /* set if wr_buffer_head > 0 or
280 dirty|flushing caps */
Sage Weilde576062009-10-06 11:31:07 -0700281 unsigned i_snap_caps; /* cap bits for snapped files */
282
283 int i_nr_by_mode[CEPH_FILE_MODE_NUM]; /* open file counts */
284
285 u32 i_truncate_seq; /* last truncate to smaller size */
286 u64 i_truncate_size; /* and the size we last truncated down to */
287 int i_truncate_pending; /* still need to call vmtruncate */
288
289 u64 i_max_size; /* max file size authorized by mds */
290 u64 i_reported_size; /* (max_)size reported to or requested of mds */
291 u64 i_wanted_max_size; /* offset we'd like to write too */
292 u64 i_requested_max_size; /* max_size we've requested */
293
294 /* held references to caps */
295 int i_pin_ref;
Henry C Changd3d07202011-05-11 10:29:54 +0000296 int i_rd_ref, i_rdcache_ref, i_wr_ref, i_wb_ref;
Sage Weilde576062009-10-06 11:31:07 -0700297 int i_wrbuffer_ref, i_wrbuffer_ref_head;
298 u32 i_shared_gen; /* increment each time we get FILE_SHARED */
Sage Weilcd045cb2010-11-04 11:05:05 -0700299 u32 i_rdcache_gen; /* incremented each time we get FILE_CACHE. */
Sage Weilde576062009-10-06 11:31:07 -0700300 u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */
301
302 struct list_head i_unsafe_writes; /* uncommitted sync writes */
303 struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */
304 spinlock_t i_unsafe_lock;
305
306 struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */
307 int i_snap_realm_counter; /* snap realm (if caps) */
308 struct list_head i_snap_realm_item;
309 struct list_head i_snap_flush_item;
310
311 struct work_struct i_wb_work; /* writeback work */
312 struct work_struct i_pg_inv_work; /* page invalidation work */
313
314 struct work_struct i_vmtruncate_work;
315
316 struct inode vfs_inode; /* at end */
317};
318
319static inline struct ceph_inode_info *ceph_inode(struct inode *inode)
320{
Noah Watkinsfbbccec2009-10-28 11:54:49 -0700321 return container_of(inode, struct ceph_inode_info, vfs_inode);
Sage Weilde576062009-10-06 11:31:07 -0700322}
323
Yehuda Sadehad1fee92011-01-21 16:44:03 -0800324static inline struct ceph_fs_client *ceph_inode_to_client(struct inode *inode)
325{
326 return (struct ceph_fs_client *)inode->i_sb->s_fs_info;
327}
328
329static inline struct ceph_fs_client *ceph_sb_to_client(struct super_block *sb)
330{
331 return (struct ceph_fs_client *)sb->s_fs_info;
332}
333
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700334static inline struct ceph_vino ceph_vino(struct inode *inode)
335{
336 return ceph_inode(inode)->i_vino;
337}
338
339/*
340 * ino_t is <64 bits on many architectures, blech.
341 *
Yehuda Sadehad1fee92011-01-21 16:44:03 -0800342 * i_ino (kernel inode) st_ino (userspace)
343 * i386 32 32
344 * x86_64+ino32 64 32
345 * x86_64 64 64
346 */
347static inline u32 ceph_ino_to_ino32(ino_t ino)
348{
349 ino ^= ino >> (sizeof(ino) * 8 - 32);
350 if (!ino)
351 ino = 1;
352 return ino;
353}
354
355/*
356 * kernel i_ino value
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700357 */
358static inline ino_t ceph_vino_to_ino(struct ceph_vino vino)
359{
360 ino_t ino = (ino_t)vino.ino; /* ^ (vino.snap << 20); */
361#if BITS_PER_LONG == 32
Yehuda Sadehad1fee92011-01-21 16:44:03 -0800362 ino = ceph_ino_to_ino32(ino);
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700363#endif
364 return ino;
365}
366
Yehuda Sadehad1fee92011-01-21 16:44:03 -0800367/*
368 * user-visible ino (stat, filldir)
369 */
370#if BITS_PER_LONG == 32
371static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino)
372{
373 return ino;
374}
375#else
376static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino)
377{
378 if (ceph_test_mount_opt(ceph_sb_to_client(sb), INO32))
379 ino = ceph_ino_to_ino32(ino);
380 return ino;
381}
382#endif
383
384
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700385/* for printf-style formatting */
386#define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap
387
388static inline u64 ceph_ino(struct inode *inode)
389{
390 return ceph_inode(inode)->i_vino.ino;
391}
392static inline u64 ceph_snap(struct inode *inode)
393{
394 return ceph_inode(inode)->i_vino.snap;
395}
396
397static inline int ceph_ino_compare(struct inode *inode, void *data)
398{
399 struct ceph_vino *pvino = (struct ceph_vino *)data;
400 struct ceph_inode_info *ci = ceph_inode(inode);
401 return ci->i_vino.ino == pvino->ino &&
402 ci->i_vino.snap == pvino->snap;
403}
404
405static inline struct inode *ceph_find_inode(struct super_block *sb,
406 struct ceph_vino vino)
407{
408 ino_t t = ceph_vino_to_ino(vino);
409 return ilookup5(sb, t, ceph_ino_compare, &vino);
410}
411
412
413/*
414 * Ceph inode.
415 */
416#define CEPH_I_COMPLETE 1 /* we have complete directory cached */
417#define CEPH_I_NODELAY 4 /* do not delay cap release */
418#define CEPH_I_FLUSH 8 /* do not delay flush of dirty metadata */
419#define CEPH_I_NOFLUSH 16 /* do not flush dirty caps */
420
Sage Weilde576062009-10-06 11:31:07 -0700421static inline void ceph_i_clear(struct inode *inode, unsigned mask)
422{
423 struct ceph_inode_info *ci = ceph_inode(inode);
424
425 spin_lock(&inode->i_lock);
426 ci->i_ceph_flags &= ~mask;
427 spin_unlock(&inode->i_lock);
428}
429
430static inline void ceph_i_set(struct inode *inode, unsigned mask)
431{
432 struct ceph_inode_info *ci = ceph_inode(inode);
433
434 spin_lock(&inode->i_lock);
435 ci->i_ceph_flags |= mask;
436 spin_unlock(&inode->i_lock);
437}
438
439static inline bool ceph_i_test(struct inode *inode, unsigned mask)
440{
441 struct ceph_inode_info *ci = ceph_inode(inode);
442 bool r;
443
Sage Weilefa4c122010-10-18 14:04:31 -0700444 spin_lock(&inode->i_lock);
Sage Weilde576062009-10-06 11:31:07 -0700445 r = (ci->i_ceph_flags & mask) == mask;
Sage Weilefa4c122010-10-18 14:04:31 -0700446 spin_unlock(&inode->i_lock);
Sage Weilde576062009-10-06 11:31:07 -0700447 return r;
448}
449
450
451/* find a specific frag @f */
452extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci,
453 u32 f);
454
455/*
456 * choose fragment for value @v. copy frag content to pfrag, if leaf
457 * exists
458 */
459extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v,
460 struct ceph_inode_frag *pfrag,
461 int *found);
462
Sage Weilde576062009-10-06 11:31:07 -0700463static inline struct ceph_dentry_info *ceph_dentry(struct dentry *dentry)
464{
465 return (struct ceph_dentry_info *)dentry->d_fsdata;
466}
467
468static inline loff_t ceph_make_fpos(unsigned frag, unsigned off)
469{
470 return ((loff_t)frag << 32) | (loff_t)off;
471}
472
Sage Weilde576062009-10-06 11:31:07 -0700473/*
474 * caps helpers
475 */
476static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci)
477{
478 return !RB_EMPTY_ROOT(&ci->i_caps);
479}
480
481extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented);
482extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t);
483extern int __ceph_caps_issued_other(struct ceph_inode_info *ci,
484 struct ceph_cap *cap);
485
486static inline int ceph_caps_issued(struct ceph_inode_info *ci)
487{
488 int issued;
489 spin_lock(&ci->vfs_inode.i_lock);
490 issued = __ceph_caps_issued(ci, NULL);
491 spin_unlock(&ci->vfs_inode.i_lock);
492 return issued;
493}
494
495static inline int ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask,
496 int touch)
497{
498 int r;
499 spin_lock(&ci->vfs_inode.i_lock);
500 r = __ceph_caps_issued_mask(ci, mask, touch);
501 spin_unlock(&ci->vfs_inode.i_lock);
502 return r;
503}
504
505static inline int __ceph_caps_dirty(struct ceph_inode_info *ci)
506{
507 return ci->i_dirty_caps | ci->i_flushing_caps;
508}
Sage Weilfca65b42011-05-04 11:33:47 -0700509extern int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask);
Sage Weilde576062009-10-06 11:31:07 -0700510
511extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask);
512extern int __ceph_caps_used(struct ceph_inode_info *ci);
513
514extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci);
515
516/*
517 * wanted, by virtue of open file modes AND cap refs (buffered/cached data)
518 */
519static inline int __ceph_caps_wanted(struct ceph_inode_info *ci)
520{
521 int w = __ceph_caps_file_wanted(ci) | __ceph_caps_used(ci);
522 if (w & CEPH_CAP_FILE_BUFFER)
523 w |= CEPH_CAP_FILE_EXCL; /* we want EXCL if dirty data */
524 return w;
525}
526
527/* what the mds thinks we want */
528extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci);
529
Yehuda Sadeh37151662010-06-17 16:16:12 -0700530extern void ceph_caps_init(struct ceph_mds_client *mdsc);
531extern void ceph_caps_finalize(struct ceph_mds_client *mdsc);
532extern void ceph_adjust_min_caps(struct ceph_mds_client *mdsc, int delta);
533extern int ceph_reserve_caps(struct ceph_mds_client *mdsc,
534 struct ceph_cap_reservation *ctx, int need);
535extern int ceph_unreserve_caps(struct ceph_mds_client *mdsc,
536 struct ceph_cap_reservation *ctx);
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700537extern void ceph_reservation_status(struct ceph_fs_client *client,
Sage Weilde576062009-10-06 11:31:07 -0700538 int *total, int *avail, int *used,
Sage Weil85ccce42010-02-17 10:02:43 -0800539 int *reserved, int *min);
Sage Weilde576062009-10-06 11:31:07 -0700540
Sage Weilde576062009-10-06 11:31:07 -0700541
Sage Weilde576062009-10-06 11:31:07 -0700542
543/*
544 * we keep buffered readdir results attached to file->private_data
545 */
Sage Weil4918b6d2011-07-26 11:26:07 -0700546#define CEPH_F_SYNC 1
547
Sage Weilde576062009-10-06 11:31:07 -0700548struct ceph_file_info {
Sage Weil252c6722011-07-26 11:25:27 -0700549 short fmode; /* initialized on open */
550 short flags; /* CEPH_F_* */
Sage Weilde576062009-10-06 11:31:07 -0700551
552 /* readdir: position within the dir */
553 u32 frag;
554 struct ceph_mds_request *last_readdir;
555 int at_end;
556
557 /* readdir: position within a frag */
558 unsigned offset; /* offset of last chunk, adjusted for . and .. */
559 u64 next_offset; /* offset of next chunk (last_name's + 1) */
560 char *last_name; /* last entry in previous chunk */
561 struct dentry *dentry; /* next dentry (for dcache readdir) */
562 unsigned long dir_release_count;
563
564 /* used for -o dirstat read() on directory thing */
565 char *dir_info;
566 int dir_info_len;
567};
568
569
570
571/*
Sage Weilde576062009-10-06 11:31:07 -0700572 * A "snap realm" describes a subset of the file hierarchy sharing
573 * the same set of snapshots that apply to it. The realms themselves
574 * are organized into a hierarchy, such that children inherit (some of)
575 * the snapshots of their parents.
576 *
577 * All inodes within the realm that have capabilities are linked into a
578 * per-realm list.
579 */
580struct ceph_snap_realm {
581 u64 ino;
582 atomic_t nref;
Sage Weila105f002010-02-15 14:37:55 -0800583 struct rb_node node;
584
Sage Weilde576062009-10-06 11:31:07 -0700585 u64 created, seq;
586 u64 parent_ino;
587 u64 parent_since; /* snapid when our current parent became so */
588
589 u64 *prior_parent_snaps; /* snaps inherited from any parents we */
590 int num_prior_parent_snaps; /* had prior to parent_since */
591 u64 *snaps; /* snaps specific to this realm */
592 int num_snaps;
593
594 struct ceph_snap_realm *parent;
595 struct list_head children; /* list of child realms */
596 struct list_head child_item;
597
598 struct list_head empty_item; /* if i have ref==0 */
599
Sage Weilae00d4f2010-09-16 16:26:51 -0700600 struct list_head dirty_item; /* if realm needs new context */
601
Sage Weilde576062009-10-06 11:31:07 -0700602 /* the current set of snaps for this realm */
603 struct ceph_snap_context *cached_context;
604
605 struct list_head inodes_with_caps;
606 spinlock_t inodes_with_caps_lock;
607};
608
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700609static inline int default_congestion_kb(void)
Sage Weilde576062009-10-06 11:31:07 -0700610{
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700611 int congestion_kb;
612
613 /*
614 * Copied from NFS
615 *
616 * congestion size, scale with available memory.
617 *
618 * 64MB: 8192k
619 * 128MB: 11585k
620 * 256MB: 16384k
621 * 512MB: 23170k
622 * 1GB: 32768k
623 * 2GB: 46340k
624 * 4GB: 65536k
625 * 8GB: 92681k
626 * 16GB: 131072k
627 *
628 * This allows larger machines to have larger/more transfers.
629 * Limit the default to 256M
630 */
631 congestion_kb = (16*int_sqrt(totalram_pages)) << (PAGE_SHIFT-10);
632 if (congestion_kb > 256*1024)
633 congestion_kb = 256*1024;
634
635 return congestion_kb;
Sage Weilde576062009-10-06 11:31:07 -0700636}
637
638
639
640/* snap.c */
641struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc,
642 u64 ino);
643extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc,
644 struct ceph_snap_realm *realm);
645extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc,
646 struct ceph_snap_realm *realm);
647extern int ceph_update_snap_trace(struct ceph_mds_client *m,
648 void *p, void *e, bool deletion);
649extern void ceph_handle_snap(struct ceph_mds_client *mdsc,
Sage Weil2600d2d2010-02-22 15:12:16 -0800650 struct ceph_mds_session *session,
Sage Weilde576062009-10-06 11:31:07 -0700651 struct ceph_msg *msg);
Sage Weilfc837c8f2010-04-13 11:41:22 -0700652extern void ceph_queue_cap_snap(struct ceph_inode_info *ci);
Sage Weilde576062009-10-06 11:31:07 -0700653extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci,
654 struct ceph_cap_snap *capsnap);
655extern void ceph_cleanup_empty_realms(struct ceph_mds_client *mdsc);
656
657/*
658 * a cap_snap is "pending" if it is still awaiting an in-progress
659 * sync write (that may/may not still update size, mtime, etc.).
660 */
661static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci)
662{
663 return !list_empty(&ci->i_cap_snaps) &&
664 list_entry(ci->i_cap_snaps.prev, struct ceph_cap_snap,
665 ci_item)->writing;
666}
667
Sage Weilde576062009-10-06 11:31:07 -0700668/* inode.c */
669extern const struct inode_operations ceph_file_iops;
670
671extern struct inode *ceph_alloc_inode(struct super_block *sb);
672extern void ceph_destroy_inode(struct inode *inode);
673
674extern struct inode *ceph_get_inode(struct super_block *sb,
675 struct ceph_vino vino);
676extern struct inode *ceph_get_snapdir(struct inode *parent);
677extern int ceph_fill_file_size(struct inode *inode, int issued,
678 u32 truncate_seq, u64 truncate_size, u64 size);
679extern void ceph_fill_file_time(struct inode *inode, int issued,
680 u64 time_warp_seq, struct timespec *ctime,
681 struct timespec *mtime, struct timespec *atime);
682extern int ceph_fill_trace(struct super_block *sb,
683 struct ceph_mds_request *req,
684 struct ceph_mds_session *session);
685extern int ceph_readdir_prepopulate(struct ceph_mds_request *req,
686 struct ceph_mds_session *session);
687
688extern int ceph_inode_holds_cap(struct inode *inode, int mask);
689
690extern int ceph_inode_set_size(struct inode *inode, loff_t size);
Sage Weilde576062009-10-06 11:31:07 -0700691extern void __ceph_do_pending_vmtruncate(struct inode *inode);
Sage Weil3c6f6b72010-02-09 15:24:44 -0800692extern void ceph_queue_vmtruncate(struct inode *inode);
693
694extern void ceph_queue_invalidate(struct inode *inode);
695extern void ceph_queue_writeback(struct inode *inode);
Sage Weilde576062009-10-06 11:31:07 -0700696
697extern int ceph_do_getattr(struct inode *inode, int mask);
Nick Pigginb74c79e2011-01-07 17:49:58 +1100698extern int ceph_permission(struct inode *inode, int mask, unsigned int flags);
Sage Weilde576062009-10-06 11:31:07 -0700699extern int ceph_setattr(struct dentry *dentry, struct iattr *attr);
700extern int ceph_getattr(struct vfsmount *mnt, struct dentry *dentry,
701 struct kstat *stat);
702
703/* xattr.c */
704extern int ceph_setxattr(struct dentry *, const char *, const void *,
705 size_t, int);
706extern ssize_t ceph_getxattr(struct dentry *, const char *, void *, size_t);
707extern ssize_t ceph_listxattr(struct dentry *, char *, size_t);
708extern int ceph_removexattr(struct dentry *, const char *);
709extern void __ceph_build_xattrs_blob(struct ceph_inode_info *ci);
710extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci);
711
712/* caps.c */
713extern const char *ceph_cap_string(int c);
714extern void ceph_handle_caps(struct ceph_mds_session *session,
715 struct ceph_msg *msg);
716extern int ceph_add_cap(struct inode *inode,
717 struct ceph_mds_session *session, u64 cap_id,
718 int fmode, unsigned issued, unsigned wanted,
719 unsigned cap, unsigned seq, u64 realmino, int flags,
720 struct ceph_cap_reservation *caps_reservation);
Sage Weil7c1332b2010-02-16 11:39:45 -0800721extern void __ceph_remove_cap(struct ceph_cap *cap);
Sage Weilde576062009-10-06 11:31:07 -0700722static inline void ceph_remove_cap(struct ceph_cap *cap)
723{
724 struct inode *inode = &cap->ci->vfs_inode;
725 spin_lock(&inode->i_lock);
Sage Weil7c1332b2010-02-16 11:39:45 -0800726 __ceph_remove_cap(cap);
Sage Weilde576062009-10-06 11:31:07 -0700727 spin_unlock(&inode->i_lock);
728}
Yehuda Sadeh37151662010-06-17 16:16:12 -0700729extern void ceph_put_cap(struct ceph_mds_client *mdsc,
730 struct ceph_cap *cap);
Sage Weilde576062009-10-06 11:31:07 -0700731
732extern void ceph_queue_caps_release(struct inode *inode);
Stephen Rothwellf1a3d572010-01-18 11:53:08 +1100733extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc);
Christoph Hellwig7ea80852010-05-26 17:53:25 +0200734extern int ceph_fsync(struct file *file, int datasync);
Sage Weilde576062009-10-06 11:31:07 -0700735extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc,
736 struct ceph_mds_session *session);
Greg Farnum2bc50252010-06-30 12:44:34 -0700737extern struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci,
738 int mds);
Sage Weilde576062009-10-06 11:31:07 -0700739extern int ceph_get_cap_mds(struct inode *inode);
740extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps);
741extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had);
742extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
743 struct ceph_snap_context *snapc);
744extern void __ceph_flush_snaps(struct ceph_inode_info *ci,
Sage Weile8351242010-09-17 08:03:08 -0700745 struct ceph_mds_session **psession,
746 int again);
Sage Weilde576062009-10-06 11:31:07 -0700747extern void ceph_check_caps(struct ceph_inode_info *ci, int flags,
748 struct ceph_mds_session *session);
Sage Weilafcdaea2009-10-14 14:27:38 -0700749extern void ceph_check_delayed_caps(struct ceph_mds_client *mdsc);
750extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc);
Sage Weilde576062009-10-06 11:31:07 -0700751
752extern int ceph_encode_inode_release(void **p, struct inode *inode,
753 int mds, int drop, int unless, int force);
754extern int ceph_encode_dentry_release(void **p, struct dentry *dn,
755 int mds, int drop, int unless);
756
757extern int ceph_get_caps(struct ceph_inode_info *ci, int need, int want,
758 int *got, loff_t endoff);
759
760/* for counting open files by mode */
761static inline void __ceph_get_fmode(struct ceph_inode_info *ci, int mode)
762{
763 ci->i_nr_by_mode[mode]++;
764}
765extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode);
766
767/* addr.c */
768extern const struct address_space_operations ceph_aops;
769extern int ceph_mmap(struct file *file, struct vm_area_struct *vma);
770
771/* file.c */
772extern const struct file_operations ceph_file_fops;
773extern const struct address_space_operations ceph_aops;
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700774extern int ceph_copy_to_page_vector(struct page **pages,
775 const char *data,
776 loff_t off, size_t len);
777extern int ceph_copy_from_page_vector(struct page **pages,
778 char *data,
779 loff_t off, size_t len);
780extern struct page **ceph_alloc_page_vector(int num_pages, gfp_t flags);
Sage Weilde576062009-10-06 11:31:07 -0700781extern int ceph_open(struct inode *inode, struct file *file);
782extern struct dentry *ceph_lookup_open(struct inode *dir, struct dentry *dentry,
783 struct nameidata *nd, int mode,
784 int locked_dir);
785extern int ceph_release(struct inode *inode, struct file *filp);
Sage Weilde576062009-10-06 11:31:07 -0700786
787/* dir.c */
788extern const struct file_operations ceph_dir_fops;
789extern const struct inode_operations ceph_dir_iops;
Sage Weil52dfb8a2010-08-03 10:25:30 -0700790extern const struct dentry_operations ceph_dentry_ops, ceph_snap_dentry_ops,
Sage Weilde576062009-10-06 11:31:07 -0700791 ceph_snapdir_dentry_ops;
792
793extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry);
794extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req,
795 struct dentry *dentry, int err);
796
797extern void ceph_dentry_lru_add(struct dentry *dn);
798extern void ceph_dentry_lru_touch(struct dentry *dn);
799extern void ceph_dentry_lru_del(struct dentry *dn);
Sage Weil81a6cf22010-05-14 09:35:38 -0700800extern void ceph_invalidate_dentry_lease(struct dentry *dentry);
Sage Weil6c0f3af2010-11-16 11:14:34 -0800801extern unsigned ceph_dentry_hash(struct dentry *dn);
Sage Weilde576062009-10-06 11:31:07 -0700802
803/*
804 * our d_ops vary depending on whether the inode is live,
805 * snapshotted (read-only), or a virtual ".snap" directory.
806 */
807int ceph_init_dentry(struct dentry *dentry);
808
809
810/* ioctl.c */
811extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
812
813/* export.c */
814extern const struct export_operations ceph_export_ops;
815
Greg Farnum40819f62010-08-02 15:34:23 -0700816/* locks.c */
817extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl);
818extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl);
819extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num);
820extern int ceph_encode_locks(struct inode *i, struct ceph_pagelist *p,
821 int p_locks, int f_locks);
822extern int lock_to_ceph_filelock(struct file_lock *fl, struct ceph_filelock *c);
823
Sage Weilde576062009-10-06 11:31:07 -0700824static inline struct inode *get_dentry_parent_inode(struct dentry *dentry)
825{
826 if (dentry && dentry->d_parent)
827 return dentry->d_parent->d_inode;
828
829 return NULL;
830}
831
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700832/* debugfs.c */
833extern int ceph_fs_debugfs_init(struct ceph_fs_client *client);
834extern void ceph_fs_debugfs_cleanup(struct ceph_fs_client *client);
835
Sage Weilde576062009-10-06 11:31:07 -0700836#endif /* _FS_CEPH_SUPER_H */