blob: 66ebe720e40dd40189510a1da9ce62f45385653b [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 Weila40dc6c2012-01-10 09:12:55 -080031#define CEPH_MOUNT_OPT_DCACHE (1<<9) /* use dcache for readdir etc */
Sage Weil6a259382010-07-07 09:06:08 -070032
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070033#define CEPH_MOUNT_OPT_DEFAULT (CEPH_MOUNT_OPT_RBYTES)
Sage Weilde576062009-10-06 11:31:07 -070034
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070035#define ceph_set_mount_opt(fsc, opt) \
36 (fsc)->mount_options->flags |= CEPH_MOUNT_OPT_##opt;
37#define ceph_test_mount_opt(fsc, opt) \
38 (!!((fsc)->mount_options->flags & CEPH_MOUNT_OPT_##opt))
Sage Weilde576062009-10-06 11:31:07 -070039
Sage Weil83817e32011-08-04 08:03:44 -070040#define CEPH_RSIZE_DEFAULT 0 /* max read size */
41#define CEPH_RASIZE_DEFAULT (8192*1024) /* readahead */
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070042#define CEPH_MAX_READDIR_DEFAULT 1024
43#define CEPH_MAX_READDIR_BYTES_DEFAULT (512*1024)
44#define CEPH_SNAPDIRNAME_DEFAULT ".snap"
Sage Weilde576062009-10-06 11:31:07 -070045
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070046struct ceph_mount_options {
Sage Weilde576062009-10-06 11:31:07 -070047 int flags;
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070048 int sb_flags;
49
Sage Weil83817e32011-08-04 08:03:44 -070050 int wsize; /* max write size */
51 int rsize; /* max read size */
52 int rasize; /* max readahead */
Sage Weil6e19a162010-04-29 16:38:32 -070053 int congestion_kb; /* max writeback in flight */
54 int caps_wanted_delay_min, caps_wanted_delay_max;
55 int cap_release_safety;
Sage Weil23804d92010-05-14 13:06:30 -070056 int max_readdir; /* max readdir result (entires) */
57 int max_readdir_bytes; /* max readdir result (bytes) */
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070058
59 /*
60 * everything above this point can be memcmp'd; everything below
61 * is handled in compare_mount_options()
62 */
63
Sage Weilde576062009-10-06 11:31:07 -070064 char *snapdir_name; /* default ".snap" */
Sage Weilde576062009-10-06 11:31:07 -070065};
66
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070067struct ceph_fs_client {
Sage Weilde576062009-10-06 11:31:07 -070068 struct super_block *sb;
69
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070070 struct ceph_mount_options *mount_options;
71 struct ceph_client *client;
72
Sage Weilde576062009-10-06 11:31:07 -070073 unsigned long mount_state;
Sage Weil85ccce42010-02-17 10:02:43 -080074 int min_caps; /* min caps i added */
75
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070076 struct ceph_mds_client *mdsc;
Sage Weilde576062009-10-06 11:31:07 -070077
78 /* writeback */
79 mempool_t *wb_pagevec_pool;
80 struct workqueue_struct *wb_wq;
81 struct workqueue_struct *pg_inv_wq;
82 struct workqueue_struct *trunc_wq;
Yehuda Sadeh2baba252009-12-18 13:51:57 -080083 atomic_long_t writeback_count;
Sage Weilde576062009-10-06 11:31:07 -070084
85 struct backing_dev_info backing_dev_info;
Sage Weil07433042009-11-18 16:50:41 -080086
87#ifdef CONFIG_DEBUG_FS
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070088 struct dentry *debugfs_dentry_lru, *debugfs_caps;
Yehuda Sadeh2baba252009-12-18 13:51:57 -080089 struct dentry *debugfs_congestion_kb;
Sage Weil06edf042009-12-15 14:44:32 -080090 struct dentry *debugfs_bdi;
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070091 struct dentry *debugfs_mdsc, *debugfs_mdsmap;
Sage Weil07433042009-11-18 16:50:41 -080092#endif
Sage Weilde576062009-10-06 11:31:07 -070093};
94
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070095
Sage Weilde576062009-10-06 11:31:07 -070096/*
97 * File i/o capability. This tracks shared state with the metadata
98 * server that allows us to cache or writeback attributes or to read
99 * and write data. For any given inode, we should have one or more
100 * capabilities, one issued by each metadata server, and our
101 * cumulative access is the OR of all issued capabilities.
102 *
103 * Each cap is referenced by the inode's i_caps rbtree and by per-mds
104 * session capability lists.
105 */
106struct ceph_cap {
107 struct ceph_inode_info *ci;
108 struct rb_node ci_node; /* per-ci cap tree */
109 struct ceph_mds_session *session;
110 struct list_head session_caps; /* per-session caplist */
111 int mds;
112 u64 cap_id; /* unique cap id (mds provided) */
113 int issued; /* latest, from the mds */
114 int implemented; /* implemented superset of issued (for revocation) */
115 int mds_wanted;
Sage Weil685f9a5d2009-11-09 12:05:48 -0800116 u32 seq, issue_seq, mseq;
117 u32 cap_gen; /* active/stale cycle */
Sage Weilde576062009-10-06 11:31:07 -0700118 unsigned long last_used;
119 struct list_head caps_item;
120};
121
122#define CHECK_CAPS_NODELAY 1 /* do not delay any further */
123#define CHECK_CAPS_AUTHONLY 2 /* only check auth cap */
124#define CHECK_CAPS_FLUSH 4 /* flush any dirty caps */
125
126/*
127 * Snapped cap state that is pending flush to mds. When a snapshot occurs,
128 * we first complete any in-process sync writes and writeback any dirty
129 * data before flushing the snapped state (tracked here) back to the MDS.
130 */
131struct ceph_cap_snap {
132 atomic_t nref;
133 struct ceph_inode_info *ci;
134 struct list_head ci_item, flushing_item;
135
136 u64 follows, flush_tid;
137 int issued, dirty;
138 struct ceph_snap_context *context;
139
Al Viro5706b272011-07-26 04:52:22 -0400140 umode_t mode;
Sage Weilde576062009-10-06 11:31:07 -0700141 uid_t uid;
142 gid_t gid;
143
Sage Weil4a625be2010-08-22 15:03:56 -0700144 struct ceph_buffer *xattr_blob;
Sage Weilde576062009-10-06 11:31:07 -0700145 u64 xattr_version;
146
147 u64 size;
148 struct timespec mtime, atime, ctime;
149 u64 time_warp_seq;
150 int writing; /* a sync write is still in progress */
151 int dirty_pages; /* dirty pages awaiting writeback */
152};
153
154static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap)
155{
Sage Weil4a625be2010-08-22 15:03:56 -0700156 if (atomic_dec_and_test(&capsnap->nref)) {
157 if (capsnap->xattr_blob)
158 ceph_buffer_put(capsnap->xattr_blob);
Sage Weilde576062009-10-06 11:31:07 -0700159 kfree(capsnap);
Sage Weil4a625be2010-08-22 15:03:56 -0700160 }
Sage Weilde576062009-10-06 11:31:07 -0700161}
162
163/*
164 * The frag tree describes how a directory is fragmented, potentially across
165 * multiple metadata servers. It is also used to indicate points where
166 * metadata authority is delegated, and whether/where metadata is replicated.
167 *
168 * A _leaf_ frag will be present in the i_fragtree IFF there is
169 * delegation info. That is, if mds >= 0 || ndist > 0.
170 */
171#define CEPH_MAX_DIRFRAG_REP 4
172
173struct ceph_inode_frag {
174 struct rb_node node;
175
176 /* fragtree state */
177 u32 frag;
178 int split_by; /* i.e. 2^(split_by) children */
179
180 /* delegation and replication info */
181 int mds; /* -1 if same authority as parent */
182 int ndist; /* >0 if replicated */
183 int dist[CEPH_MAX_DIRFRAG_REP];
184};
185
186/*
187 * We cache inode xattrs as an encoded blob until they are first used,
188 * at which point we parse them into an rbtree.
189 */
190struct ceph_inode_xattr {
191 struct rb_node node;
192
193 const char *name;
194 int name_len;
195 const char *val;
196 int val_len;
197 int dirty;
198
199 int should_free_name;
200 int should_free_val;
201};
202
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700203/*
204 * Ceph dentry state
205 */
206struct ceph_dentry_info {
Sage Weilb58dc412011-03-15 15:53:40 -0700207 unsigned long flags;
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700208 struct ceph_mds_session *lease_session;
209 u32 lease_gen, lease_shared_gen;
210 u32 lease_seq;
211 unsigned long lease_renew_after, lease_renew_from;
212 struct list_head lru;
213 struct dentry *dentry;
214 u64 time;
215 u64 offset;
216};
217
Sage Weilb58dc412011-03-15 15:53:40 -0700218/*
219 * dentry flags
220 *
221 * The locking for D_COMPLETE is a bit odd:
222 * - we can clear it at almost any time (see ceph_d_prune)
223 * - it is only meaningful if:
Sage Weilbe655592011-11-30 09:47:09 -0800224 * - we hold dir inode i_ceph_lock
Sage Weilb58dc412011-03-15 15:53:40 -0700225 * - we hold dir FILE_SHARED caps
226 * - the dentry D_COMPLETE is set
227 */
228#define CEPH_D_COMPLETE 1 /* if set, d_u.d_subdirs is complete directory */
229
Sage Weilde576062009-10-06 11:31:07 -0700230struct ceph_inode_xattrs_info {
231 /*
232 * (still encoded) xattr blob. we avoid the overhead of parsing
233 * this until someone actually calls getxattr, etc.
234 *
235 * blob->vec.iov_len == 4 implies there are no xattrs; blob ==
236 * NULL means we don't know.
237 */
238 struct ceph_buffer *blob, *prealloc_blob;
239
240 struct rb_root index;
241 bool dirty;
242 int count;
243 int names_size;
244 int vals_size;
245 u64 version, index_version;
246};
247
248/*
249 * Ceph inode.
250 */
Sage Weilde576062009-10-06 11:31:07 -0700251struct ceph_inode_info {
252 struct ceph_vino i_vino; /* ceph ino + snap */
253
Sage Weilbe655592011-11-30 09:47:09 -0800254 spinlock_t i_ceph_lock;
255
Sage Weilde576062009-10-06 11:31:07 -0700256 u64 i_version;
257 u32 i_time_warp_seq;
258
259 unsigned i_ceph_flags;
260 unsigned long i_release_count;
261
Sage Weil6c0f3af2010-11-16 11:14:34 -0800262 struct ceph_dir_layout i_dir_layout;
Sage Weilde576062009-10-06 11:31:07 -0700263 struct ceph_file_layout i_layout;
264 char *i_symlink;
265
266 /* for dirs */
267 struct timespec i_rctime;
268 u64 i_rbytes, i_rfiles, i_rsubdirs;
269 u64 i_files, i_subdirs;
Sage Weilc6ffe102011-11-03 09:23:36 -0700270 u64 i_max_offset; /* largest readdir offset, set with D_COMPLETE */
Sage Weilde576062009-10-06 11:31:07 -0700271
272 struct rb_root i_fragtree;
273 struct mutex i_fragtree_mutex;
274
275 struct ceph_inode_xattrs_info i_xattrs;
276
Sage Weilbe655592011-11-30 09:47:09 -0800277 /* capabilities. protected _both_ by i_ceph_lock and cap->session's
Sage Weilde576062009-10-06 11:31:07 -0700278 * s_mutex. */
279 struct rb_root i_caps; /* cap list */
280 struct ceph_cap *i_auth_cap; /* authoritative cap, if any */
281 unsigned i_dirty_caps, i_flushing_caps; /* mask of dirtied fields */
282 struct list_head i_dirty_item, i_flushing_item;
283 u64 i_cap_flush_seq;
284 /* we need to track cap writeback on a per-cap-bit basis, to allow
285 * overlapping, pipelined cap flushes to the mds. we can probably
286 * reduce the tid to 8 bits if we're concerned about inode size. */
287 u16 i_cap_flush_last_tid, i_cap_flush_tid[CEPH_CAP_BITS];
288 wait_queue_head_t i_cap_wq; /* threads waiting on a capability */
289 unsigned long i_hold_caps_min; /* jiffies */
290 unsigned long i_hold_caps_max; /* jiffies */
291 struct list_head i_cap_delay_list; /* for delayed cap release to mds */
292 int i_cap_exporting_mds; /* to handle cap migration between */
293 unsigned i_cap_exporting_mseq; /* mds's. */
294 unsigned i_cap_exporting_issued;
295 struct ceph_cap_reservation i_cap_migration_resv;
296 struct list_head i_cap_snaps; /* snapped state pending flush to mds */
Sage Weil7d8cb262010-08-24 08:44:16 -0700297 struct ceph_snap_context *i_head_snapc; /* set if wr_buffer_head > 0 or
298 dirty|flushing caps */
Sage Weilde576062009-10-06 11:31:07 -0700299 unsigned i_snap_caps; /* cap bits for snapped files */
300
301 int i_nr_by_mode[CEPH_FILE_MODE_NUM]; /* open file counts */
302
303 u32 i_truncate_seq; /* last truncate to smaller size */
304 u64 i_truncate_size; /* and the size we last truncated down to */
305 int i_truncate_pending; /* still need to call vmtruncate */
306
307 u64 i_max_size; /* max file size authorized by mds */
308 u64 i_reported_size; /* (max_)size reported to or requested of mds */
309 u64 i_wanted_max_size; /* offset we'd like to write too */
310 u64 i_requested_max_size; /* max_size we've requested */
311
312 /* held references to caps */
313 int i_pin_ref;
Henry C Changd3d07202011-05-11 10:29:54 +0000314 int i_rd_ref, i_rdcache_ref, i_wr_ref, i_wb_ref;
Sage Weilde576062009-10-06 11:31:07 -0700315 int i_wrbuffer_ref, i_wrbuffer_ref_head;
316 u32 i_shared_gen; /* increment each time we get FILE_SHARED */
Sage Weilcd045cb2010-11-04 11:05:05 -0700317 u32 i_rdcache_gen; /* incremented each time we get FILE_CACHE. */
Sage Weilde576062009-10-06 11:31:07 -0700318 u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */
319
320 struct list_head i_unsafe_writes; /* uncommitted sync writes */
321 struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */
322 spinlock_t i_unsafe_lock;
323
324 struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */
325 int i_snap_realm_counter; /* snap realm (if caps) */
326 struct list_head i_snap_realm_item;
327 struct list_head i_snap_flush_item;
328
329 struct work_struct i_wb_work; /* writeback work */
330 struct work_struct i_pg_inv_work; /* page invalidation work */
331
332 struct work_struct i_vmtruncate_work;
333
334 struct inode vfs_inode; /* at end */
335};
336
337static inline struct ceph_inode_info *ceph_inode(struct inode *inode)
338{
Noah Watkinsfbbccec2009-10-28 11:54:49 -0700339 return container_of(inode, struct ceph_inode_info, vfs_inode);
Sage Weilde576062009-10-06 11:31:07 -0700340}
341
Yehuda Sadehad1fee92011-01-21 16:44:03 -0800342static inline struct ceph_fs_client *ceph_inode_to_client(struct inode *inode)
343{
344 return (struct ceph_fs_client *)inode->i_sb->s_fs_info;
345}
346
347static inline struct ceph_fs_client *ceph_sb_to_client(struct super_block *sb)
348{
349 return (struct ceph_fs_client *)sb->s_fs_info;
350}
351
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700352static inline struct ceph_vino ceph_vino(struct inode *inode)
353{
354 return ceph_inode(inode)->i_vino;
355}
356
357/*
358 * ino_t is <64 bits on many architectures, blech.
359 *
Yehuda Sadehad1fee92011-01-21 16:44:03 -0800360 * i_ino (kernel inode) st_ino (userspace)
361 * i386 32 32
362 * x86_64+ino32 64 32
363 * x86_64 64 64
364 */
Amon Ott3310f752011-10-20 13:04:07 -0700365static inline u32 ceph_ino_to_ino32(__u64 vino)
Yehuda Sadehad1fee92011-01-21 16:44:03 -0800366{
Amon Ott3310f752011-10-20 13:04:07 -0700367 u32 ino = vino & 0xffffffff;
368 ino ^= vino >> 32;
Yehuda Sadehad1fee92011-01-21 16:44:03 -0800369 if (!ino)
Amon Otta661fc52012-01-23 09:25:23 -0800370 ino = 2;
Yehuda Sadehad1fee92011-01-21 16:44:03 -0800371 return ino;
372}
373
374/*
375 * kernel i_ino value
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700376 */
377static inline ino_t ceph_vino_to_ino(struct ceph_vino vino)
378{
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700379#if BITS_PER_LONG == 32
Amon Ott3310f752011-10-20 13:04:07 -0700380 return ceph_ino_to_ino32(vino.ino);
381#else
382 return (ino_t)vino.ino;
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700383#endif
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700384}
385
Yehuda Sadehad1fee92011-01-21 16:44:03 -0800386/*
387 * user-visible ino (stat, filldir)
388 */
389#if BITS_PER_LONG == 32
390static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino)
391{
392 return ino;
393}
394#else
395static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino)
396{
397 if (ceph_test_mount_opt(ceph_sb_to_client(sb), INO32))
398 ino = ceph_ino_to_ino32(ino);
399 return ino;
400}
401#endif
402
403
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700404/* for printf-style formatting */
405#define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap
406
407static inline u64 ceph_ino(struct inode *inode)
408{
409 return ceph_inode(inode)->i_vino.ino;
410}
411static inline u64 ceph_snap(struct inode *inode)
412{
413 return ceph_inode(inode)->i_vino.snap;
414}
415
416static inline int ceph_ino_compare(struct inode *inode, void *data)
417{
418 struct ceph_vino *pvino = (struct ceph_vino *)data;
419 struct ceph_inode_info *ci = ceph_inode(inode);
420 return ci->i_vino.ino == pvino->ino &&
421 ci->i_vino.snap == pvino->snap;
422}
423
424static inline struct inode *ceph_find_inode(struct super_block *sb,
425 struct ceph_vino vino)
426{
427 ino_t t = ceph_vino_to_ino(vino);
428 return ilookup5(sb, t, ceph_ino_compare, &vino);
429}
430
431
432/*
433 * Ceph inode.
434 */
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700435#define CEPH_I_NODELAY 4 /* do not delay cap release */
436#define CEPH_I_FLUSH 8 /* do not delay flush of dirty metadata */
437#define CEPH_I_NOFLUSH 16 /* do not flush dirty caps */
438
Sage Weilde576062009-10-06 11:31:07 -0700439static inline void ceph_i_clear(struct inode *inode, unsigned mask)
440{
441 struct ceph_inode_info *ci = ceph_inode(inode);
442
Sage Weilbe655592011-11-30 09:47:09 -0800443 spin_lock(&ci->i_ceph_lock);
Sage Weilde576062009-10-06 11:31:07 -0700444 ci->i_ceph_flags &= ~mask;
Sage Weilbe655592011-11-30 09:47:09 -0800445 spin_unlock(&ci->i_ceph_lock);
Sage Weilde576062009-10-06 11:31:07 -0700446}
447
448static inline void ceph_i_set(struct inode *inode, unsigned mask)
449{
450 struct ceph_inode_info *ci = ceph_inode(inode);
451
Sage Weilbe655592011-11-30 09:47:09 -0800452 spin_lock(&ci->i_ceph_lock);
Sage Weilde576062009-10-06 11:31:07 -0700453 ci->i_ceph_flags |= mask;
Sage Weilbe655592011-11-30 09:47:09 -0800454 spin_unlock(&ci->i_ceph_lock);
Sage Weilde576062009-10-06 11:31:07 -0700455}
456
457static inline bool ceph_i_test(struct inode *inode, unsigned mask)
458{
459 struct ceph_inode_info *ci = ceph_inode(inode);
460 bool r;
461
Sage Weilbe655592011-11-30 09:47:09 -0800462 spin_lock(&ci->i_ceph_lock);
Sage Weilde576062009-10-06 11:31:07 -0700463 r = (ci->i_ceph_flags & mask) == mask;
Sage Weilbe655592011-11-30 09:47:09 -0800464 spin_unlock(&ci->i_ceph_lock);
Sage Weilde576062009-10-06 11:31:07 -0700465 return r;
466}
467
468
469/* find a specific frag @f */
470extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci,
471 u32 f);
472
473/*
474 * choose fragment for value @v. copy frag content to pfrag, if leaf
475 * exists
476 */
477extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v,
478 struct ceph_inode_frag *pfrag,
479 int *found);
480
Sage Weilde576062009-10-06 11:31:07 -0700481static inline struct ceph_dentry_info *ceph_dentry(struct dentry *dentry)
482{
483 return (struct ceph_dentry_info *)dentry->d_fsdata;
484}
485
486static inline loff_t ceph_make_fpos(unsigned frag, unsigned off)
487{
488 return ((loff_t)frag << 32) | (loff_t)off;
489}
490
Sage Weilde576062009-10-06 11:31:07 -0700491/*
Sage Weilc6ffe102011-11-03 09:23:36 -0700492 * set/clear directory D_COMPLETE flag
493 */
494void ceph_dir_set_complete(struct inode *inode);
495void ceph_dir_clear_complete(struct inode *inode);
496bool ceph_dir_test_complete(struct inode *inode);
497
498/*
Sage Weilde576062009-10-06 11:31:07 -0700499 * caps helpers
500 */
501static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci)
502{
503 return !RB_EMPTY_ROOT(&ci->i_caps);
504}
505
506extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented);
507extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t);
508extern int __ceph_caps_issued_other(struct ceph_inode_info *ci,
509 struct ceph_cap *cap);
510
511static inline int ceph_caps_issued(struct ceph_inode_info *ci)
512{
513 int issued;
Sage Weilbe655592011-11-30 09:47:09 -0800514 spin_lock(&ci->i_ceph_lock);
Sage Weilde576062009-10-06 11:31:07 -0700515 issued = __ceph_caps_issued(ci, NULL);
Sage Weilbe655592011-11-30 09:47:09 -0800516 spin_unlock(&ci->i_ceph_lock);
Sage Weilde576062009-10-06 11:31:07 -0700517 return issued;
518}
519
520static inline int ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask,
521 int touch)
522{
523 int r;
Sage Weilbe655592011-11-30 09:47:09 -0800524 spin_lock(&ci->i_ceph_lock);
Sage Weilde576062009-10-06 11:31:07 -0700525 r = __ceph_caps_issued_mask(ci, mask, touch);
Sage Weilbe655592011-11-30 09:47:09 -0800526 spin_unlock(&ci->i_ceph_lock);
Sage Weilde576062009-10-06 11:31:07 -0700527 return r;
528}
529
530static inline int __ceph_caps_dirty(struct ceph_inode_info *ci)
531{
532 return ci->i_dirty_caps | ci->i_flushing_caps;
533}
Sage Weilfca65b42011-05-04 11:33:47 -0700534extern int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask);
Sage Weilde576062009-10-06 11:31:07 -0700535
536extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask);
537extern int __ceph_caps_used(struct ceph_inode_info *ci);
538
539extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci);
540
541/*
542 * wanted, by virtue of open file modes AND cap refs (buffered/cached data)
543 */
544static inline int __ceph_caps_wanted(struct ceph_inode_info *ci)
545{
546 int w = __ceph_caps_file_wanted(ci) | __ceph_caps_used(ci);
547 if (w & CEPH_CAP_FILE_BUFFER)
548 w |= CEPH_CAP_FILE_EXCL; /* we want EXCL if dirty data */
549 return w;
550}
551
552/* what the mds thinks we want */
553extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci);
554
Yehuda Sadeh37151662010-06-17 16:16:12 -0700555extern void ceph_caps_init(struct ceph_mds_client *mdsc);
556extern void ceph_caps_finalize(struct ceph_mds_client *mdsc);
557extern void ceph_adjust_min_caps(struct ceph_mds_client *mdsc, int delta);
558extern int ceph_reserve_caps(struct ceph_mds_client *mdsc,
559 struct ceph_cap_reservation *ctx, int need);
560extern int ceph_unreserve_caps(struct ceph_mds_client *mdsc,
561 struct ceph_cap_reservation *ctx);
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700562extern void ceph_reservation_status(struct ceph_fs_client *client,
Sage Weilde576062009-10-06 11:31:07 -0700563 int *total, int *avail, int *used,
Sage Weil85ccce42010-02-17 10:02:43 -0800564 int *reserved, int *min);
Sage Weilde576062009-10-06 11:31:07 -0700565
Sage Weilde576062009-10-06 11:31:07 -0700566
Sage Weilde576062009-10-06 11:31:07 -0700567
568/*
569 * we keep buffered readdir results attached to file->private_data
570 */
Sage Weil4918b6d2011-07-26 11:26:07 -0700571#define CEPH_F_SYNC 1
Sage Weil9cfa1092011-07-26 11:26:18 -0700572#define CEPH_F_ATEND 2
Sage Weil4918b6d2011-07-26 11:26:07 -0700573
Sage Weilde576062009-10-06 11:31:07 -0700574struct ceph_file_info {
Sage Weil252c6722011-07-26 11:25:27 -0700575 short fmode; /* initialized on open */
576 short flags; /* CEPH_F_* */
Sage Weilde576062009-10-06 11:31:07 -0700577
578 /* readdir: position within the dir */
579 u32 frag;
580 struct ceph_mds_request *last_readdir;
Sage Weilde576062009-10-06 11:31:07 -0700581
582 /* readdir: position within a frag */
583 unsigned offset; /* offset of last chunk, adjusted for . and .. */
584 u64 next_offset; /* offset of next chunk (last_name's + 1) */
585 char *last_name; /* last entry in previous chunk */
586 struct dentry *dentry; /* next dentry (for dcache readdir) */
587 unsigned long dir_release_count;
588
589 /* used for -o dirstat read() on directory thing */
590 char *dir_info;
591 int dir_info_len;
592};
593
594
595
596/*
Sage Weilde576062009-10-06 11:31:07 -0700597 * A "snap realm" describes a subset of the file hierarchy sharing
598 * the same set of snapshots that apply to it. The realms themselves
599 * are organized into a hierarchy, such that children inherit (some of)
600 * the snapshots of their parents.
601 *
602 * All inodes within the realm that have capabilities are linked into a
603 * per-realm list.
604 */
605struct ceph_snap_realm {
606 u64 ino;
607 atomic_t nref;
Sage Weila105f002010-02-15 14:37:55 -0800608 struct rb_node node;
609
Sage Weilde576062009-10-06 11:31:07 -0700610 u64 created, seq;
611 u64 parent_ino;
612 u64 parent_since; /* snapid when our current parent became so */
613
614 u64 *prior_parent_snaps; /* snaps inherited from any parents we */
Alex Elderaa711ee2012-07-13 20:35:11 -0500615 u32 num_prior_parent_snaps; /* had prior to parent_since */
Sage Weilde576062009-10-06 11:31:07 -0700616 u64 *snaps; /* snaps specific to this realm */
Alex Elderaa711ee2012-07-13 20:35:11 -0500617 u32 num_snaps;
Sage Weilde576062009-10-06 11:31:07 -0700618
619 struct ceph_snap_realm *parent;
620 struct list_head children; /* list of child realms */
621 struct list_head child_item;
622
623 struct list_head empty_item; /* if i have ref==0 */
624
Sage Weilae00d4f2010-09-16 16:26:51 -0700625 struct list_head dirty_item; /* if realm needs new context */
626
Sage Weilde576062009-10-06 11:31:07 -0700627 /* the current set of snaps for this realm */
628 struct ceph_snap_context *cached_context;
629
630 struct list_head inodes_with_caps;
631 spinlock_t inodes_with_caps_lock;
632};
633
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700634static inline int default_congestion_kb(void)
Sage Weilde576062009-10-06 11:31:07 -0700635{
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700636 int congestion_kb;
637
638 /*
639 * Copied from NFS
640 *
641 * congestion size, scale with available memory.
642 *
643 * 64MB: 8192k
644 * 128MB: 11585k
645 * 256MB: 16384k
646 * 512MB: 23170k
647 * 1GB: 32768k
648 * 2GB: 46340k
649 * 4GB: 65536k
650 * 8GB: 92681k
651 * 16GB: 131072k
652 *
653 * This allows larger machines to have larger/more transfers.
654 * Limit the default to 256M
655 */
656 congestion_kb = (16*int_sqrt(totalram_pages)) << (PAGE_SHIFT-10);
657 if (congestion_kb > 256*1024)
658 congestion_kb = 256*1024;
659
660 return congestion_kb;
Sage Weilde576062009-10-06 11:31:07 -0700661}
662
663
664
665/* snap.c */
666struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc,
667 u64 ino);
668extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc,
669 struct ceph_snap_realm *realm);
670extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc,
671 struct ceph_snap_realm *realm);
672extern int ceph_update_snap_trace(struct ceph_mds_client *m,
673 void *p, void *e, bool deletion);
674extern void ceph_handle_snap(struct ceph_mds_client *mdsc,
Sage Weil2600d2d2010-02-22 15:12:16 -0800675 struct ceph_mds_session *session,
Sage Weilde576062009-10-06 11:31:07 -0700676 struct ceph_msg *msg);
Sage Weilfc837c8f2010-04-13 11:41:22 -0700677extern void ceph_queue_cap_snap(struct ceph_inode_info *ci);
Sage Weilde576062009-10-06 11:31:07 -0700678extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci,
679 struct ceph_cap_snap *capsnap);
680extern void ceph_cleanup_empty_realms(struct ceph_mds_client *mdsc);
681
682/*
683 * a cap_snap is "pending" if it is still awaiting an in-progress
684 * sync write (that may/may not still update size, mtime, etc.).
685 */
686static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci)
687{
688 return !list_empty(&ci->i_cap_snaps) &&
689 list_entry(ci->i_cap_snaps.prev, struct ceph_cap_snap,
690 ci_item)->writing;
691}
692
Sage Weilde576062009-10-06 11:31:07 -0700693/* inode.c */
694extern const struct inode_operations ceph_file_iops;
695
696extern struct inode *ceph_alloc_inode(struct super_block *sb);
697extern void ceph_destroy_inode(struct inode *inode);
698
699extern struct inode *ceph_get_inode(struct super_block *sb,
700 struct ceph_vino vino);
701extern struct inode *ceph_get_snapdir(struct inode *parent);
702extern int ceph_fill_file_size(struct inode *inode, int issued,
703 u32 truncate_seq, u64 truncate_size, u64 size);
704extern void ceph_fill_file_time(struct inode *inode, int issued,
705 u64 time_warp_seq, struct timespec *ctime,
706 struct timespec *mtime, struct timespec *atime);
707extern int ceph_fill_trace(struct super_block *sb,
708 struct ceph_mds_request *req,
709 struct ceph_mds_session *session);
710extern int ceph_readdir_prepopulate(struct ceph_mds_request *req,
711 struct ceph_mds_session *session);
712
713extern int ceph_inode_holds_cap(struct inode *inode, int mask);
714
715extern int ceph_inode_set_size(struct inode *inode, loff_t size);
Sage Weilde576062009-10-06 11:31:07 -0700716extern void __ceph_do_pending_vmtruncate(struct inode *inode);
Sage Weil3c6f6b72010-02-09 15:24:44 -0800717extern void ceph_queue_vmtruncate(struct inode *inode);
718
719extern void ceph_queue_invalidate(struct inode *inode);
720extern void ceph_queue_writeback(struct inode *inode);
Sage Weilde576062009-10-06 11:31:07 -0700721
722extern int ceph_do_getattr(struct inode *inode, int mask);
Al Viro10556cb2011-06-20 19:28:19 -0400723extern int ceph_permission(struct inode *inode, int mask);
Sage Weilde576062009-10-06 11:31:07 -0700724extern int ceph_setattr(struct dentry *dentry, struct iattr *attr);
725extern int ceph_getattr(struct vfsmount *mnt, struct dentry *dentry,
726 struct kstat *stat);
727
728/* xattr.c */
729extern int ceph_setxattr(struct dentry *, const char *, const void *,
730 size_t, int);
731extern ssize_t ceph_getxattr(struct dentry *, const char *, void *, size_t);
732extern ssize_t ceph_listxattr(struct dentry *, char *, size_t);
733extern int ceph_removexattr(struct dentry *, const char *);
734extern void __ceph_build_xattrs_blob(struct ceph_inode_info *ci);
735extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci);
Alex Elder3ce6cd12012-01-23 15:49:28 -0600736extern void __init ceph_xattr_init(void);
737extern void ceph_xattr_exit(void);
Sage Weilde576062009-10-06 11:31:07 -0700738
739/* caps.c */
740extern const char *ceph_cap_string(int c);
741extern void ceph_handle_caps(struct ceph_mds_session *session,
742 struct ceph_msg *msg);
743extern int ceph_add_cap(struct inode *inode,
744 struct ceph_mds_session *session, u64 cap_id,
745 int fmode, unsigned issued, unsigned wanted,
746 unsigned cap, unsigned seq, u64 realmino, int flags,
747 struct ceph_cap_reservation *caps_reservation);
Sage Weil7c1332b2010-02-16 11:39:45 -0800748extern void __ceph_remove_cap(struct ceph_cap *cap);
Sage Weilde576062009-10-06 11:31:07 -0700749static inline void ceph_remove_cap(struct ceph_cap *cap)
750{
Sage Weilbe655592011-11-30 09:47:09 -0800751 spin_lock(&cap->ci->i_ceph_lock);
Sage Weil7c1332b2010-02-16 11:39:45 -0800752 __ceph_remove_cap(cap);
Sage Weilbe655592011-11-30 09:47:09 -0800753 spin_unlock(&cap->ci->i_ceph_lock);
Sage Weilde576062009-10-06 11:31:07 -0700754}
Yehuda Sadeh37151662010-06-17 16:16:12 -0700755extern void ceph_put_cap(struct ceph_mds_client *mdsc,
756 struct ceph_cap *cap);
Sage Weilde576062009-10-06 11:31:07 -0700757
758extern void ceph_queue_caps_release(struct inode *inode);
Stephen Rothwellf1a3d572010-01-18 11:53:08 +1100759extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc);
Josef Bacik02c24a82011-07-16 20:44:56 -0400760extern int ceph_fsync(struct file *file, loff_t start, loff_t end,
761 int datasync);
Sage Weilde576062009-10-06 11:31:07 -0700762extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc,
763 struct ceph_mds_session *session);
Greg Farnum2bc50252010-06-30 12:44:34 -0700764extern struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci,
765 int mds);
Sage Weilde576062009-10-06 11:31:07 -0700766extern int ceph_get_cap_mds(struct inode *inode);
767extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps);
768extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had);
769extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
770 struct ceph_snap_context *snapc);
771extern void __ceph_flush_snaps(struct ceph_inode_info *ci,
Sage Weile8351242010-09-17 08:03:08 -0700772 struct ceph_mds_session **psession,
773 int again);
Sage Weilde576062009-10-06 11:31:07 -0700774extern void ceph_check_caps(struct ceph_inode_info *ci, int flags,
775 struct ceph_mds_session *session);
Sage Weilafcdaea2009-10-14 14:27:38 -0700776extern void ceph_check_delayed_caps(struct ceph_mds_client *mdsc);
777extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc);
Sage Weilde576062009-10-06 11:31:07 -0700778
779extern int ceph_encode_inode_release(void **p, struct inode *inode,
780 int mds, int drop, int unless, int force);
781extern int ceph_encode_dentry_release(void **p, struct dentry *dn,
782 int mds, int drop, int unless);
783
784extern int ceph_get_caps(struct ceph_inode_info *ci, int need, int want,
785 int *got, loff_t endoff);
786
787/* for counting open files by mode */
788static inline void __ceph_get_fmode(struct ceph_inode_info *ci, int mode)
789{
790 ci->i_nr_by_mode[mode]++;
791}
792extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode);
793
794/* addr.c */
795extern const struct address_space_operations ceph_aops;
796extern int ceph_mmap(struct file *file, struct vm_area_struct *vma);
797
798/* file.c */
799extern const struct file_operations ceph_file_fops;
800extern const struct address_space_operations ceph_aops;
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700801extern int ceph_copy_to_page_vector(struct page **pages,
802 const char *data,
803 loff_t off, size_t len);
804extern int ceph_copy_from_page_vector(struct page **pages,
805 char *data,
806 loff_t off, size_t len);
807extern struct page **ceph_alloc_page_vector(int num_pages, gfp_t flags);
Sage Weilde576062009-10-06 11:31:07 -0700808extern int ceph_open(struct inode *inode, struct file *file);
Sage Weil5ef50c32012-07-31 11:27:36 -0700809extern int ceph_atomic_open(struct inode *dir, struct dentry *dentry,
810 struct file *file, unsigned flags, umode_t mode,
811 int *opened);
Sage Weilde576062009-10-06 11:31:07 -0700812extern int ceph_release(struct inode *inode, struct file *filp);
Sage Weilde576062009-10-06 11:31:07 -0700813
814/* dir.c */
815extern const struct file_operations ceph_dir_fops;
816extern const struct inode_operations ceph_dir_iops;
Sage Weil52dfb8a2010-08-03 10:25:30 -0700817extern const struct dentry_operations ceph_dentry_ops, ceph_snap_dentry_ops,
Sage Weilde576062009-10-06 11:31:07 -0700818 ceph_snapdir_dentry_ops;
819
820extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry);
Sage Weil468640e2011-07-26 11:28:11 -0700821extern int ceph_handle_snapdir(struct ceph_mds_request *req,
822 struct dentry *dentry, int err);
Sage Weilde576062009-10-06 11:31:07 -0700823extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req,
824 struct dentry *dentry, int err);
825
826extern void ceph_dentry_lru_add(struct dentry *dn);
827extern void ceph_dentry_lru_touch(struct dentry *dn);
828extern void ceph_dentry_lru_del(struct dentry *dn);
Sage Weil81a6cf22010-05-14 09:35:38 -0700829extern void ceph_invalidate_dentry_lease(struct dentry *dentry);
Sage Weile5f86dc2011-07-26 11:30:55 -0700830extern unsigned ceph_dentry_hash(struct inode *dir, struct dentry *dn);
Sage Weil5f21c962011-07-26 11:30:29 -0700831extern struct inode *ceph_get_dentry_parent_inode(struct dentry *dentry);
Sage Weilde576062009-10-06 11:31:07 -0700832
833/*
834 * our d_ops vary depending on whether the inode is live,
835 * snapshotted (read-only), or a virtual ".snap" directory.
836 */
837int ceph_init_dentry(struct dentry *dentry);
838
839
840/* ioctl.c */
841extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
842
843/* export.c */
844extern const struct export_operations ceph_export_ops;
845
Greg Farnum40819f62010-08-02 15:34:23 -0700846/* locks.c */
847extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl);
848extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl);
849extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num);
850extern int ceph_encode_locks(struct inode *i, struct ceph_pagelist *p,
851 int p_locks, int f_locks);
852extern int lock_to_ceph_filelock(struct file_lock *fl, struct ceph_filelock *c);
853
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700854/* debugfs.c */
855extern int ceph_fs_debugfs_init(struct ceph_fs_client *client);
856extern void ceph_fs_debugfs_cleanup(struct ceph_fs_client *client);
857
Sage Weilde576062009-10-06 11:31:07 -0700858#endif /* _FS_CEPH_SUPER_H */