blob: 5c72430b8f7171abbebe4c203f2f7a8d0ebed030 [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 Weil83817e32011-08-04 08:03:44 -070039#define CEPH_RSIZE_DEFAULT 0 /* max read size */
40#define CEPH_RASIZE_DEFAULT (8192*1024) /* readahead */
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070041#define CEPH_MAX_READDIR_DEFAULT 1024
42#define CEPH_MAX_READDIR_BYTES_DEFAULT (512*1024)
43#define CEPH_SNAPDIRNAME_DEFAULT ".snap"
Sage Weilde576062009-10-06 11:31:07 -070044
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070045struct ceph_mount_options {
Sage Weilde576062009-10-06 11:31:07 -070046 int flags;
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070047 int sb_flags;
48
Sage Weil83817e32011-08-04 08:03:44 -070049 int wsize; /* max write size */
50 int rsize; /* max read size */
51 int rasize; /* max readahead */
Sage Weil6e19a162010-04-29 16:38:32 -070052 int congestion_kb; /* max writeback in flight */
53 int caps_wanted_delay_min, caps_wanted_delay_max;
54 int cap_release_safety;
Sage Weil23804d92010-05-14 13:06:30 -070055 int max_readdir; /* max readdir result (entires) */
56 int max_readdir_bytes; /* max readdir result (bytes) */
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070057
58 /*
59 * everything above this point can be memcmp'd; everything below
60 * is handled in compare_mount_options()
61 */
62
Sage Weilde576062009-10-06 11:31:07 -070063 char *snapdir_name; /* default ".snap" */
Sage Weilde576062009-10-06 11:31:07 -070064};
65
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070066struct ceph_fs_client {
Sage Weilde576062009-10-06 11:31:07 -070067 struct super_block *sb;
68
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070069 struct ceph_mount_options *mount_options;
70 struct ceph_client *client;
71
Sage Weilde576062009-10-06 11:31:07 -070072 unsigned long mount_state;
Sage Weil85ccce42010-02-17 10:02:43 -080073 int min_caps; /* min caps i added */
74
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070075 struct ceph_mds_client *mdsc;
Sage Weilde576062009-10-06 11:31:07 -070076
77 /* writeback */
78 mempool_t *wb_pagevec_pool;
79 struct workqueue_struct *wb_wq;
80 struct workqueue_struct *pg_inv_wq;
81 struct workqueue_struct *trunc_wq;
Yehuda Sadeh2baba252009-12-18 13:51:57 -080082 atomic_long_t writeback_count;
Sage Weilde576062009-10-06 11:31:07 -070083
84 struct backing_dev_info backing_dev_info;
Sage Weil07433042009-11-18 16:50:41 -080085
86#ifdef CONFIG_DEBUG_FS
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070087 struct dentry *debugfs_dentry_lru, *debugfs_caps;
Yehuda Sadeh2baba252009-12-18 13:51:57 -080088 struct dentry *debugfs_congestion_kb;
Sage Weil06edf042009-12-15 14:44:32 -080089 struct dentry *debugfs_bdi;
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070090 struct dentry *debugfs_mdsc, *debugfs_mdsmap;
Sage Weil07433042009-11-18 16:50:41 -080091#endif
Sage Weilde576062009-10-06 11:31:07 -070092};
93
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070094
Sage Weilde576062009-10-06 11:31:07 -070095/*
96 * File i/o capability. This tracks shared state with the metadata
97 * server that allows us to cache or writeback attributes or to read
98 * and write data. For any given inode, we should have one or more
99 * capabilities, one issued by each metadata server, and our
100 * cumulative access is the OR of all issued capabilities.
101 *
102 * Each cap is referenced by the inode's i_caps rbtree and by per-mds
103 * session capability lists.
104 */
105struct ceph_cap {
106 struct ceph_inode_info *ci;
107 struct rb_node ci_node; /* per-ci cap tree */
108 struct ceph_mds_session *session;
109 struct list_head session_caps; /* per-session caplist */
110 int mds;
111 u64 cap_id; /* unique cap id (mds provided) */
112 int issued; /* latest, from the mds */
113 int implemented; /* implemented superset of issued (for revocation) */
114 int mds_wanted;
Sage Weil685f9a5d2009-11-09 12:05:48 -0800115 u32 seq, issue_seq, mseq;
116 u32 cap_gen; /* active/stale cycle */
Sage Weilde576062009-10-06 11:31:07 -0700117 unsigned long last_used;
118 struct list_head caps_item;
119};
120
121#define CHECK_CAPS_NODELAY 1 /* do not delay any further */
122#define CHECK_CAPS_AUTHONLY 2 /* only check auth cap */
123#define CHECK_CAPS_FLUSH 4 /* flush any dirty caps */
124
125/*
126 * Snapped cap state that is pending flush to mds. When a snapshot occurs,
127 * we first complete any in-process sync writes and writeback any dirty
128 * data before flushing the snapped state (tracked here) back to the MDS.
129 */
130struct ceph_cap_snap {
131 atomic_t nref;
132 struct ceph_inode_info *ci;
133 struct list_head ci_item, flushing_item;
134
135 u64 follows, flush_tid;
136 int issued, dirty;
137 struct ceph_snap_context *context;
138
139 mode_t mode;
140 uid_t uid;
141 gid_t gid;
142
Sage Weil4a625be2010-08-22 15:03:56 -0700143 struct ceph_buffer *xattr_blob;
Sage Weilde576062009-10-06 11:31:07 -0700144 u64 xattr_version;
145
146 u64 size;
147 struct timespec mtime, atime, ctime;
148 u64 time_warp_seq;
149 int writing; /* a sync write is still in progress */
150 int dirty_pages; /* dirty pages awaiting writeback */
151};
152
153static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap)
154{
Sage Weil4a625be2010-08-22 15:03:56 -0700155 if (atomic_dec_and_test(&capsnap->nref)) {
156 if (capsnap->xattr_blob)
157 ceph_buffer_put(capsnap->xattr_blob);
Sage Weilde576062009-10-06 11:31:07 -0700158 kfree(capsnap);
Sage Weil4a625be2010-08-22 15:03:56 -0700159 }
Sage Weilde576062009-10-06 11:31:07 -0700160}
161
162/*
163 * The frag tree describes how a directory is fragmented, potentially across
164 * multiple metadata servers. It is also used to indicate points where
165 * metadata authority is delegated, and whether/where metadata is replicated.
166 *
167 * A _leaf_ frag will be present in the i_fragtree IFF there is
168 * delegation info. That is, if mds >= 0 || ndist > 0.
169 */
170#define CEPH_MAX_DIRFRAG_REP 4
171
172struct ceph_inode_frag {
173 struct rb_node node;
174
175 /* fragtree state */
176 u32 frag;
177 int split_by; /* i.e. 2^(split_by) children */
178
179 /* delegation and replication info */
180 int mds; /* -1 if same authority as parent */
181 int ndist; /* >0 if replicated */
182 int dist[CEPH_MAX_DIRFRAG_REP];
183};
184
185/*
186 * We cache inode xattrs as an encoded blob until they are first used,
187 * at which point we parse them into an rbtree.
188 */
189struct ceph_inode_xattr {
190 struct rb_node node;
191
192 const char *name;
193 int name_len;
194 const char *val;
195 int val_len;
196 int dirty;
197
198 int should_free_name;
199 int should_free_val;
200};
201
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700202/*
203 * Ceph dentry state
204 */
205struct ceph_dentry_info {
206 struct ceph_mds_session *lease_session;
207 u32 lease_gen, lease_shared_gen;
208 u32 lease_seq;
209 unsigned long lease_renew_after, lease_renew_from;
210 struct list_head lru;
211 struct dentry *dentry;
212 u64 time;
213 u64 offset;
214};
215
Sage Weilde576062009-10-06 11:31:07 -0700216struct ceph_inode_xattrs_info {
217 /*
218 * (still encoded) xattr blob. we avoid the overhead of parsing
219 * this until someone actually calls getxattr, etc.
220 *
221 * blob->vec.iov_len == 4 implies there are no xattrs; blob ==
222 * NULL means we don't know.
223 */
224 struct ceph_buffer *blob, *prealloc_blob;
225
226 struct rb_root index;
227 bool dirty;
228 int count;
229 int names_size;
230 int vals_size;
231 u64 version, index_version;
232};
233
234/*
235 * Ceph inode.
236 */
Sage Weilde576062009-10-06 11:31:07 -0700237struct ceph_inode_info {
238 struct ceph_vino i_vino; /* ceph ino + snap */
239
240 u64 i_version;
241 u32 i_time_warp_seq;
242
243 unsigned i_ceph_flags;
244 unsigned long i_release_count;
245
Sage Weil6c0f3af2010-11-16 11:14:34 -0800246 struct ceph_dir_layout i_dir_layout;
Sage Weilde576062009-10-06 11:31:07 -0700247 struct ceph_file_layout i_layout;
248 char *i_symlink;
249
250 /* for dirs */
251 struct timespec i_rctime;
252 u64 i_rbytes, i_rfiles, i_rsubdirs;
253 u64 i_files, i_subdirs;
254 u64 i_max_offset; /* largest readdir offset, set with I_COMPLETE */
255
256 struct rb_root i_fragtree;
257 struct mutex i_fragtree_mutex;
258
259 struct ceph_inode_xattrs_info i_xattrs;
260
261 /* capabilities. protected _both_ by i_lock and cap->session's
262 * s_mutex. */
263 struct rb_root i_caps; /* cap list */
264 struct ceph_cap *i_auth_cap; /* authoritative cap, if any */
265 unsigned i_dirty_caps, i_flushing_caps; /* mask of dirtied fields */
266 struct list_head i_dirty_item, i_flushing_item;
267 u64 i_cap_flush_seq;
268 /* we need to track cap writeback on a per-cap-bit basis, to allow
269 * overlapping, pipelined cap flushes to the mds. we can probably
270 * reduce the tid to 8 bits if we're concerned about inode size. */
271 u16 i_cap_flush_last_tid, i_cap_flush_tid[CEPH_CAP_BITS];
272 wait_queue_head_t i_cap_wq; /* threads waiting on a capability */
273 unsigned long i_hold_caps_min; /* jiffies */
274 unsigned long i_hold_caps_max; /* jiffies */
275 struct list_head i_cap_delay_list; /* for delayed cap release to mds */
276 int i_cap_exporting_mds; /* to handle cap migration between */
277 unsigned i_cap_exporting_mseq; /* mds's. */
278 unsigned i_cap_exporting_issued;
279 struct ceph_cap_reservation i_cap_migration_resv;
280 struct list_head i_cap_snaps; /* snapped state pending flush to mds */
Sage Weil7d8cb262010-08-24 08:44:16 -0700281 struct ceph_snap_context *i_head_snapc; /* set if wr_buffer_head > 0 or
282 dirty|flushing caps */
Sage Weilde576062009-10-06 11:31:07 -0700283 unsigned i_snap_caps; /* cap bits for snapped files */
284
285 int i_nr_by_mode[CEPH_FILE_MODE_NUM]; /* open file counts */
286
287 u32 i_truncate_seq; /* last truncate to smaller size */
288 u64 i_truncate_size; /* and the size we last truncated down to */
289 int i_truncate_pending; /* still need to call vmtruncate */
290
291 u64 i_max_size; /* max file size authorized by mds */
292 u64 i_reported_size; /* (max_)size reported to or requested of mds */
293 u64 i_wanted_max_size; /* offset we'd like to write too */
294 u64 i_requested_max_size; /* max_size we've requested */
295
296 /* held references to caps */
297 int i_pin_ref;
Henry C Changd3d07202011-05-11 10:29:54 +0000298 int i_rd_ref, i_rdcache_ref, i_wr_ref, i_wb_ref;
Sage Weilde576062009-10-06 11:31:07 -0700299 int i_wrbuffer_ref, i_wrbuffer_ref_head;
300 u32 i_shared_gen; /* increment each time we get FILE_SHARED */
Sage Weilcd045cb2010-11-04 11:05:05 -0700301 u32 i_rdcache_gen; /* incremented each time we get FILE_CACHE. */
Sage Weilde576062009-10-06 11:31:07 -0700302 u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */
303
304 struct list_head i_unsafe_writes; /* uncommitted sync writes */
305 struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */
306 spinlock_t i_unsafe_lock;
307
308 struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */
309 int i_snap_realm_counter; /* snap realm (if caps) */
310 struct list_head i_snap_realm_item;
311 struct list_head i_snap_flush_item;
312
313 struct work_struct i_wb_work; /* writeback work */
314 struct work_struct i_pg_inv_work; /* page invalidation work */
315
316 struct work_struct i_vmtruncate_work;
317
318 struct inode vfs_inode; /* at end */
319};
320
321static inline struct ceph_inode_info *ceph_inode(struct inode *inode)
322{
Noah Watkinsfbbccec2009-10-28 11:54:49 -0700323 return container_of(inode, struct ceph_inode_info, vfs_inode);
Sage Weilde576062009-10-06 11:31:07 -0700324}
325
Yehuda Sadehad1fee92011-01-21 16:44:03 -0800326static inline struct ceph_fs_client *ceph_inode_to_client(struct inode *inode)
327{
328 return (struct ceph_fs_client *)inode->i_sb->s_fs_info;
329}
330
331static inline struct ceph_fs_client *ceph_sb_to_client(struct super_block *sb)
332{
333 return (struct ceph_fs_client *)sb->s_fs_info;
334}
335
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700336static inline struct ceph_vino ceph_vino(struct inode *inode)
337{
338 return ceph_inode(inode)->i_vino;
339}
340
341/*
342 * ino_t is <64 bits on many architectures, blech.
343 *
Yehuda Sadehad1fee92011-01-21 16:44:03 -0800344 * i_ino (kernel inode) st_ino (userspace)
345 * i386 32 32
346 * x86_64+ino32 64 32
347 * x86_64 64 64
348 */
349static inline u32 ceph_ino_to_ino32(ino_t ino)
350{
351 ino ^= ino >> (sizeof(ino) * 8 - 32);
352 if (!ino)
353 ino = 1;
354 return ino;
355}
356
357/*
358 * kernel i_ino value
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700359 */
360static inline ino_t ceph_vino_to_ino(struct ceph_vino vino)
361{
362 ino_t ino = (ino_t)vino.ino; /* ^ (vino.snap << 20); */
363#if BITS_PER_LONG == 32
Yehuda Sadehad1fee92011-01-21 16:44:03 -0800364 ino = ceph_ino_to_ino32(ino);
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700365#endif
366 return ino;
367}
368
Yehuda Sadehad1fee92011-01-21 16:44:03 -0800369/*
370 * user-visible ino (stat, filldir)
371 */
372#if BITS_PER_LONG == 32
373static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino)
374{
375 return ino;
376}
377#else
378static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino)
379{
380 if (ceph_test_mount_opt(ceph_sb_to_client(sb), INO32))
381 ino = ceph_ino_to_ino32(ino);
382 return ino;
383}
384#endif
385
386
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700387/* for printf-style formatting */
388#define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap
389
390static inline u64 ceph_ino(struct inode *inode)
391{
392 return ceph_inode(inode)->i_vino.ino;
393}
394static inline u64 ceph_snap(struct inode *inode)
395{
396 return ceph_inode(inode)->i_vino.snap;
397}
398
399static inline int ceph_ino_compare(struct inode *inode, void *data)
400{
401 struct ceph_vino *pvino = (struct ceph_vino *)data;
402 struct ceph_inode_info *ci = ceph_inode(inode);
403 return ci->i_vino.ino == pvino->ino &&
404 ci->i_vino.snap == pvino->snap;
405}
406
407static inline struct inode *ceph_find_inode(struct super_block *sb,
408 struct ceph_vino vino)
409{
410 ino_t t = ceph_vino_to_ino(vino);
411 return ilookup5(sb, t, ceph_ino_compare, &vino);
412}
413
414
415/*
416 * Ceph inode.
417 */
418#define CEPH_I_COMPLETE 1 /* we have complete directory cached */
419#define CEPH_I_NODELAY 4 /* do not delay cap release */
420#define CEPH_I_FLUSH 8 /* do not delay flush of dirty metadata */
421#define CEPH_I_NOFLUSH 16 /* do not flush dirty caps */
422
Sage Weilde576062009-10-06 11:31:07 -0700423static inline void ceph_i_clear(struct inode *inode, unsigned mask)
424{
425 struct ceph_inode_info *ci = ceph_inode(inode);
426
427 spin_lock(&inode->i_lock);
428 ci->i_ceph_flags &= ~mask;
429 spin_unlock(&inode->i_lock);
430}
431
432static inline void ceph_i_set(struct inode *inode, unsigned mask)
433{
434 struct ceph_inode_info *ci = ceph_inode(inode);
435
436 spin_lock(&inode->i_lock);
437 ci->i_ceph_flags |= mask;
438 spin_unlock(&inode->i_lock);
439}
440
441static inline bool ceph_i_test(struct inode *inode, unsigned mask)
442{
443 struct ceph_inode_info *ci = ceph_inode(inode);
444 bool r;
445
Sage Weilefa4c122010-10-18 14:04:31 -0700446 spin_lock(&inode->i_lock);
Sage Weilde576062009-10-06 11:31:07 -0700447 r = (ci->i_ceph_flags & mask) == mask;
Sage Weilefa4c122010-10-18 14:04:31 -0700448 spin_unlock(&inode->i_lock);
Sage Weilde576062009-10-06 11:31:07 -0700449 return r;
450}
451
452
453/* find a specific frag @f */
454extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci,
455 u32 f);
456
457/*
458 * choose fragment for value @v. copy frag content to pfrag, if leaf
459 * exists
460 */
461extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v,
462 struct ceph_inode_frag *pfrag,
463 int *found);
464
Sage Weilde576062009-10-06 11:31:07 -0700465static inline struct ceph_dentry_info *ceph_dentry(struct dentry *dentry)
466{
467 return (struct ceph_dentry_info *)dentry->d_fsdata;
468}
469
470static inline loff_t ceph_make_fpos(unsigned frag, unsigned off)
471{
472 return ((loff_t)frag << 32) | (loff_t)off;
473}
474
Sage Weilde576062009-10-06 11:31:07 -0700475/*
476 * caps helpers
477 */
478static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci)
479{
480 return !RB_EMPTY_ROOT(&ci->i_caps);
481}
482
483extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented);
484extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t);
485extern int __ceph_caps_issued_other(struct ceph_inode_info *ci,
486 struct ceph_cap *cap);
487
488static inline int ceph_caps_issued(struct ceph_inode_info *ci)
489{
490 int issued;
491 spin_lock(&ci->vfs_inode.i_lock);
492 issued = __ceph_caps_issued(ci, NULL);
493 spin_unlock(&ci->vfs_inode.i_lock);
494 return issued;
495}
496
497static inline int ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask,
498 int touch)
499{
500 int r;
501 spin_lock(&ci->vfs_inode.i_lock);
502 r = __ceph_caps_issued_mask(ci, mask, touch);
503 spin_unlock(&ci->vfs_inode.i_lock);
504 return r;
505}
506
507static inline int __ceph_caps_dirty(struct ceph_inode_info *ci)
508{
509 return ci->i_dirty_caps | ci->i_flushing_caps;
510}
Sage Weilfca65b42011-05-04 11:33:47 -0700511extern int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask);
Sage Weilde576062009-10-06 11:31:07 -0700512
513extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask);
514extern int __ceph_caps_used(struct ceph_inode_info *ci);
515
516extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci);
517
518/*
519 * wanted, by virtue of open file modes AND cap refs (buffered/cached data)
520 */
521static inline int __ceph_caps_wanted(struct ceph_inode_info *ci)
522{
523 int w = __ceph_caps_file_wanted(ci) | __ceph_caps_used(ci);
524 if (w & CEPH_CAP_FILE_BUFFER)
525 w |= CEPH_CAP_FILE_EXCL; /* we want EXCL if dirty data */
526 return w;
527}
528
529/* what the mds thinks we want */
530extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci);
531
Yehuda Sadeh37151662010-06-17 16:16:12 -0700532extern void ceph_caps_init(struct ceph_mds_client *mdsc);
533extern void ceph_caps_finalize(struct ceph_mds_client *mdsc);
534extern void ceph_adjust_min_caps(struct ceph_mds_client *mdsc, int delta);
535extern int ceph_reserve_caps(struct ceph_mds_client *mdsc,
536 struct ceph_cap_reservation *ctx, int need);
537extern int ceph_unreserve_caps(struct ceph_mds_client *mdsc,
538 struct ceph_cap_reservation *ctx);
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700539extern void ceph_reservation_status(struct ceph_fs_client *client,
Sage Weilde576062009-10-06 11:31:07 -0700540 int *total, int *avail, int *used,
Sage Weil85ccce42010-02-17 10:02:43 -0800541 int *reserved, int *min);
Sage Weilde576062009-10-06 11:31:07 -0700542
Sage Weilde576062009-10-06 11:31:07 -0700543
Sage Weilde576062009-10-06 11:31:07 -0700544
545/*
546 * we keep buffered readdir results attached to file->private_data
547 */
Sage Weil4918b6d2011-07-26 11:26:07 -0700548#define CEPH_F_SYNC 1
Sage Weil9cfa1092011-07-26 11:26:18 -0700549#define CEPH_F_ATEND 2
Sage Weil4918b6d2011-07-26 11:26:07 -0700550
Sage Weilde576062009-10-06 11:31:07 -0700551struct ceph_file_info {
Sage Weil252c6722011-07-26 11:25:27 -0700552 short fmode; /* initialized on open */
553 short flags; /* CEPH_F_* */
Sage Weilde576062009-10-06 11:31:07 -0700554
555 /* readdir: position within the dir */
556 u32 frag;
557 struct ceph_mds_request *last_readdir;
Sage Weilde576062009-10-06 11:31:07 -0700558
559 /* readdir: position within a frag */
560 unsigned offset; /* offset of last chunk, adjusted for . and .. */
561 u64 next_offset; /* offset of next chunk (last_name's + 1) */
562 char *last_name; /* last entry in previous chunk */
563 struct dentry *dentry; /* next dentry (for dcache readdir) */
564 unsigned long dir_release_count;
565
566 /* used for -o dirstat read() on directory thing */
567 char *dir_info;
568 int dir_info_len;
569};
570
571
572
573/*
Sage Weilde576062009-10-06 11:31:07 -0700574 * A "snap realm" describes a subset of the file hierarchy sharing
575 * the same set of snapshots that apply to it. The realms themselves
576 * are organized into a hierarchy, such that children inherit (some of)
577 * the snapshots of their parents.
578 *
579 * All inodes within the realm that have capabilities are linked into a
580 * per-realm list.
581 */
582struct ceph_snap_realm {
583 u64 ino;
584 atomic_t nref;
Sage Weila105f002010-02-15 14:37:55 -0800585 struct rb_node node;
586
Sage Weilde576062009-10-06 11:31:07 -0700587 u64 created, seq;
588 u64 parent_ino;
589 u64 parent_since; /* snapid when our current parent became so */
590
591 u64 *prior_parent_snaps; /* snaps inherited from any parents we */
592 int num_prior_parent_snaps; /* had prior to parent_since */
593 u64 *snaps; /* snaps specific to this realm */
594 int num_snaps;
595
596 struct ceph_snap_realm *parent;
597 struct list_head children; /* list of child realms */
598 struct list_head child_item;
599
600 struct list_head empty_item; /* if i have ref==0 */
601
Sage Weilae00d4f2010-09-16 16:26:51 -0700602 struct list_head dirty_item; /* if realm needs new context */
603
Sage Weilde576062009-10-06 11:31:07 -0700604 /* the current set of snaps for this realm */
605 struct ceph_snap_context *cached_context;
606
607 struct list_head inodes_with_caps;
608 spinlock_t inodes_with_caps_lock;
609};
610
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700611static inline int default_congestion_kb(void)
Sage Weilde576062009-10-06 11:31:07 -0700612{
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700613 int congestion_kb;
614
615 /*
616 * Copied from NFS
617 *
618 * congestion size, scale with available memory.
619 *
620 * 64MB: 8192k
621 * 128MB: 11585k
622 * 256MB: 16384k
623 * 512MB: 23170k
624 * 1GB: 32768k
625 * 2GB: 46340k
626 * 4GB: 65536k
627 * 8GB: 92681k
628 * 16GB: 131072k
629 *
630 * This allows larger machines to have larger/more transfers.
631 * Limit the default to 256M
632 */
633 congestion_kb = (16*int_sqrt(totalram_pages)) << (PAGE_SHIFT-10);
634 if (congestion_kb > 256*1024)
635 congestion_kb = 256*1024;
636
637 return congestion_kb;
Sage Weilde576062009-10-06 11:31:07 -0700638}
639
640
641
642/* snap.c */
643struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc,
644 u64 ino);
645extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc,
646 struct ceph_snap_realm *realm);
647extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc,
648 struct ceph_snap_realm *realm);
649extern int ceph_update_snap_trace(struct ceph_mds_client *m,
650 void *p, void *e, bool deletion);
651extern void ceph_handle_snap(struct ceph_mds_client *mdsc,
Sage Weil2600d2d2010-02-22 15:12:16 -0800652 struct ceph_mds_session *session,
Sage Weilde576062009-10-06 11:31:07 -0700653 struct ceph_msg *msg);
Sage Weilfc837c8f2010-04-13 11:41:22 -0700654extern void ceph_queue_cap_snap(struct ceph_inode_info *ci);
Sage Weilde576062009-10-06 11:31:07 -0700655extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci,
656 struct ceph_cap_snap *capsnap);
657extern void ceph_cleanup_empty_realms(struct ceph_mds_client *mdsc);
658
659/*
660 * a cap_snap is "pending" if it is still awaiting an in-progress
661 * sync write (that may/may not still update size, mtime, etc.).
662 */
663static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci)
664{
665 return !list_empty(&ci->i_cap_snaps) &&
666 list_entry(ci->i_cap_snaps.prev, struct ceph_cap_snap,
667 ci_item)->writing;
668}
669
Sage Weilde576062009-10-06 11:31:07 -0700670/* inode.c */
671extern const struct inode_operations ceph_file_iops;
672
673extern struct inode *ceph_alloc_inode(struct super_block *sb);
674extern void ceph_destroy_inode(struct inode *inode);
675
676extern struct inode *ceph_get_inode(struct super_block *sb,
677 struct ceph_vino vino);
678extern struct inode *ceph_get_snapdir(struct inode *parent);
679extern int ceph_fill_file_size(struct inode *inode, int issued,
680 u32 truncate_seq, u64 truncate_size, u64 size);
681extern void ceph_fill_file_time(struct inode *inode, int issued,
682 u64 time_warp_seq, struct timespec *ctime,
683 struct timespec *mtime, struct timespec *atime);
684extern int ceph_fill_trace(struct super_block *sb,
685 struct ceph_mds_request *req,
686 struct ceph_mds_session *session);
687extern int ceph_readdir_prepopulate(struct ceph_mds_request *req,
688 struct ceph_mds_session *session);
689
690extern int ceph_inode_holds_cap(struct inode *inode, int mask);
691
692extern int ceph_inode_set_size(struct inode *inode, loff_t size);
Sage Weilde576062009-10-06 11:31:07 -0700693extern void __ceph_do_pending_vmtruncate(struct inode *inode);
Sage Weil3c6f6b72010-02-09 15:24:44 -0800694extern void ceph_queue_vmtruncate(struct inode *inode);
695
696extern void ceph_queue_invalidate(struct inode *inode);
697extern void ceph_queue_writeback(struct inode *inode);
Sage Weilde576062009-10-06 11:31:07 -0700698
699extern int ceph_do_getattr(struct inode *inode, int mask);
Al Viro10556cb2011-06-20 19:28:19 -0400700extern int ceph_permission(struct inode *inode, int mask);
Sage Weilde576062009-10-06 11:31:07 -0700701extern int ceph_setattr(struct dentry *dentry, struct iattr *attr);
702extern int ceph_getattr(struct vfsmount *mnt, struct dentry *dentry,
703 struct kstat *stat);
704
705/* xattr.c */
706extern int ceph_setxattr(struct dentry *, const char *, const void *,
707 size_t, int);
708extern ssize_t ceph_getxattr(struct dentry *, const char *, void *, size_t);
709extern ssize_t ceph_listxattr(struct dentry *, char *, size_t);
710extern int ceph_removexattr(struct dentry *, const char *);
711extern void __ceph_build_xattrs_blob(struct ceph_inode_info *ci);
712extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci);
713
714/* caps.c */
715extern const char *ceph_cap_string(int c);
716extern void ceph_handle_caps(struct ceph_mds_session *session,
717 struct ceph_msg *msg);
718extern int ceph_add_cap(struct inode *inode,
719 struct ceph_mds_session *session, u64 cap_id,
720 int fmode, unsigned issued, unsigned wanted,
721 unsigned cap, unsigned seq, u64 realmino, int flags,
722 struct ceph_cap_reservation *caps_reservation);
Sage Weil7c1332b2010-02-16 11:39:45 -0800723extern void __ceph_remove_cap(struct ceph_cap *cap);
Sage Weilde576062009-10-06 11:31:07 -0700724static inline void ceph_remove_cap(struct ceph_cap *cap)
725{
726 struct inode *inode = &cap->ci->vfs_inode;
727 spin_lock(&inode->i_lock);
Sage Weil7c1332b2010-02-16 11:39:45 -0800728 __ceph_remove_cap(cap);
Sage Weilde576062009-10-06 11:31:07 -0700729 spin_unlock(&inode->i_lock);
730}
Yehuda Sadeh37151662010-06-17 16:16:12 -0700731extern void ceph_put_cap(struct ceph_mds_client *mdsc,
732 struct ceph_cap *cap);
Sage Weilde576062009-10-06 11:31:07 -0700733
734extern void ceph_queue_caps_release(struct inode *inode);
Stephen Rothwellf1a3d572010-01-18 11:53:08 +1100735extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc);
Josef Bacik02c24a82011-07-16 20:44:56 -0400736extern int ceph_fsync(struct file *file, loff_t start, loff_t end,
737 int datasync);
Sage Weilde576062009-10-06 11:31:07 -0700738extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc,
739 struct ceph_mds_session *session);
Greg Farnum2bc50252010-06-30 12:44:34 -0700740extern struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci,
741 int mds);
Sage Weilde576062009-10-06 11:31:07 -0700742extern int ceph_get_cap_mds(struct inode *inode);
743extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps);
744extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had);
745extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
746 struct ceph_snap_context *snapc);
747extern void __ceph_flush_snaps(struct ceph_inode_info *ci,
Sage Weile8351242010-09-17 08:03:08 -0700748 struct ceph_mds_session **psession,
749 int again);
Sage Weilde576062009-10-06 11:31:07 -0700750extern void ceph_check_caps(struct ceph_inode_info *ci, int flags,
751 struct ceph_mds_session *session);
Sage Weilafcdaea2009-10-14 14:27:38 -0700752extern void ceph_check_delayed_caps(struct ceph_mds_client *mdsc);
753extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc);
Sage Weilde576062009-10-06 11:31:07 -0700754
755extern int ceph_encode_inode_release(void **p, struct inode *inode,
756 int mds, int drop, int unless, int force);
757extern int ceph_encode_dentry_release(void **p, struct dentry *dn,
758 int mds, int drop, int unless);
759
760extern int ceph_get_caps(struct ceph_inode_info *ci, int need, int want,
761 int *got, loff_t endoff);
762
763/* for counting open files by mode */
764static inline void __ceph_get_fmode(struct ceph_inode_info *ci, int mode)
765{
766 ci->i_nr_by_mode[mode]++;
767}
768extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode);
769
770/* addr.c */
771extern const struct address_space_operations ceph_aops;
772extern int ceph_mmap(struct file *file, struct vm_area_struct *vma);
773
774/* file.c */
775extern const struct file_operations ceph_file_fops;
776extern const struct address_space_operations ceph_aops;
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700777extern int ceph_copy_to_page_vector(struct page **pages,
778 const char *data,
779 loff_t off, size_t len);
780extern int ceph_copy_from_page_vector(struct page **pages,
781 char *data,
782 loff_t off, size_t len);
783extern struct page **ceph_alloc_page_vector(int num_pages, gfp_t flags);
Sage Weilde576062009-10-06 11:31:07 -0700784extern int ceph_open(struct inode *inode, struct file *file);
785extern struct dentry *ceph_lookup_open(struct inode *dir, struct dentry *dentry,
786 struct nameidata *nd, int mode,
787 int locked_dir);
788extern int ceph_release(struct inode *inode, struct file *filp);
Sage Weilde576062009-10-06 11:31:07 -0700789
790/* dir.c */
791extern const struct file_operations ceph_dir_fops;
792extern const struct inode_operations ceph_dir_iops;
Sage Weil52dfb8a2010-08-03 10:25:30 -0700793extern const struct dentry_operations ceph_dentry_ops, ceph_snap_dentry_ops,
Sage Weilde576062009-10-06 11:31:07 -0700794 ceph_snapdir_dentry_ops;
795
796extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry);
Sage Weil468640e2011-07-26 11:28:11 -0700797extern int ceph_handle_snapdir(struct ceph_mds_request *req,
798 struct dentry *dentry, int err);
Sage Weilde576062009-10-06 11:31:07 -0700799extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req,
800 struct dentry *dentry, int err);
801
802extern void ceph_dentry_lru_add(struct dentry *dn);
803extern void ceph_dentry_lru_touch(struct dentry *dn);
804extern void ceph_dentry_lru_del(struct dentry *dn);
Sage Weil81a6cf22010-05-14 09:35:38 -0700805extern void ceph_invalidate_dentry_lease(struct dentry *dentry);
Sage Weile5f86dc2011-07-26 11:30:55 -0700806extern unsigned ceph_dentry_hash(struct inode *dir, struct dentry *dn);
Sage Weil5f21c962011-07-26 11:30:29 -0700807extern struct inode *ceph_get_dentry_parent_inode(struct dentry *dentry);
Sage Weilde576062009-10-06 11:31:07 -0700808
809/*
810 * our d_ops vary depending on whether the inode is live,
811 * snapshotted (read-only), or a virtual ".snap" directory.
812 */
813int ceph_init_dentry(struct dentry *dentry);
814
815
816/* ioctl.c */
817extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
818
819/* export.c */
820extern const struct export_operations ceph_export_ops;
821
Greg Farnum40819f62010-08-02 15:34:23 -0700822/* locks.c */
823extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl);
824extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl);
825extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num);
826extern int ceph_encode_locks(struct inode *i, struct ceph_pagelist *p,
827 int p_locks, int f_locks);
828extern int lock_to_ceph_filelock(struct file_lock *fl, struct ceph_filelock *c);
829
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700830/* debugfs.c */
831extern int ceph_fs_debugfs_init(struct ceph_fs_client *client);
832extern void ceph_fs_debugfs_cleanup(struct ceph_fs_client *client);
833
Sage Weilde576062009-10-06 11:31:07 -0700834#endif /* _FS_CEPH_SUPER_H */