blob: cbded572345e77a107e539aa4e433d6f6f7964c0 [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. */
Sage Weil92a49fb2013-02-22 15:31:00 -080024#define CEPH_BLOCK_SHIFT 22 /* 4 MB */
Sage Weilde576062009-10-06 11:31:07 -070025#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;
Eric W. Biederman05cb11c2013-01-31 02:56:19 -0800141 kuid_t uid;
142 kgid_t gid;
Sage Weilde576062009-10-06 11:31:07 -0700143
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 {
207 struct ceph_mds_session *lease_session;
208 u32 lease_gen, lease_shared_gen;
209 u32 lease_seq;
210 unsigned long lease_renew_after, lease_renew_from;
211 struct list_head lru;
212 struct dentry *dentry;
213 u64 time;
214 u64 offset;
215};
216
Sage Weilde576062009-10-06 11:31:07 -0700217struct ceph_inode_xattrs_info {
218 /*
219 * (still encoded) xattr blob. we avoid the overhead of parsing
220 * this until someone actually calls getxattr, etc.
221 *
222 * blob->vec.iov_len == 4 implies there are no xattrs; blob ==
223 * NULL means we don't know.
224 */
225 struct ceph_buffer *blob, *prealloc_blob;
226
227 struct rb_root index;
228 bool dirty;
229 int count;
230 int names_size;
231 int vals_size;
232 u64 version, index_version;
233};
234
235/*
236 * Ceph inode.
237 */
Sage Weilde576062009-10-06 11:31:07 -0700238struct ceph_inode_info {
239 struct ceph_vino i_vino; /* ceph ino + snap */
240
Sage Weilbe655592011-11-30 09:47:09 -0800241 spinlock_t i_ceph_lock;
242
Sage Weilde576062009-10-06 11:31:07 -0700243 u64 i_version;
244 u32 i_time_warp_seq;
245
246 unsigned i_ceph_flags;
Yan, Zheng2f276c52013-03-13 19:44:32 +0800247 atomic_t i_release_count;
248 atomic_t i_complete_count;
Sage Weilde576062009-10-06 11:31:07 -0700249
Sage Weil6c0f3af2010-11-16 11:14:34 -0800250 struct ceph_dir_layout i_dir_layout;
Sage Weilde576062009-10-06 11:31:07 -0700251 struct ceph_file_layout i_layout;
252 char *i_symlink;
253
254 /* for dirs */
255 struct timespec i_rctime;
256 u64 i_rbytes, i_rfiles, i_rsubdirs;
257 u64 i_files, i_subdirs;
Yan, Zheng2f276c52013-03-13 19:44:32 +0800258 u64 i_max_offset; /* largest readdir offset, set with complete dir */
Sage Weilde576062009-10-06 11:31:07 -0700259
260 struct rb_root i_fragtree;
261 struct mutex i_fragtree_mutex;
262
263 struct ceph_inode_xattrs_info i_xattrs;
264
Sage Weilbe655592011-11-30 09:47:09 -0800265 /* capabilities. protected _both_ by i_ceph_lock and cap->session's
Sage Weilde576062009-10-06 11:31:07 -0700266 * s_mutex. */
267 struct rb_root i_caps; /* cap list */
268 struct ceph_cap *i_auth_cap; /* authoritative cap, if any */
269 unsigned i_dirty_caps, i_flushing_caps; /* mask of dirtied fields */
270 struct list_head i_dirty_item, i_flushing_item;
271 u64 i_cap_flush_seq;
272 /* we need to track cap writeback on a per-cap-bit basis, to allow
273 * overlapping, pipelined cap flushes to the mds. we can probably
274 * reduce the tid to 8 bits if we're concerned about inode size. */
275 u16 i_cap_flush_last_tid, i_cap_flush_tid[CEPH_CAP_BITS];
276 wait_queue_head_t i_cap_wq; /* threads waiting on a capability */
277 unsigned long i_hold_caps_min; /* jiffies */
278 unsigned long i_hold_caps_max; /* jiffies */
279 struct list_head i_cap_delay_list; /* for delayed cap release to mds */
280 int i_cap_exporting_mds; /* to handle cap migration between */
281 unsigned i_cap_exporting_mseq; /* mds's. */
282 unsigned i_cap_exporting_issued;
283 struct ceph_cap_reservation i_cap_migration_resv;
284 struct list_head i_cap_snaps; /* snapped state pending flush to mds */
Sage Weil7d8cb262010-08-24 08:44:16 -0700285 struct ceph_snap_context *i_head_snapc; /* set if wr_buffer_head > 0 or
286 dirty|flushing caps */
Sage Weilde576062009-10-06 11:31:07 -0700287 unsigned i_snap_caps; /* cap bits for snapped files */
288
289 int i_nr_by_mode[CEPH_FILE_MODE_NUM]; /* open file counts */
290
291 u32 i_truncate_seq; /* last truncate to smaller size */
292 u64 i_truncate_size; /* and the size we last truncated down to */
293 int i_truncate_pending; /* still need to call vmtruncate */
294
295 u64 i_max_size; /* max file size authorized by mds */
296 u64 i_reported_size; /* (max_)size reported to or requested of mds */
297 u64 i_wanted_max_size; /* offset we'd like to write too */
298 u64 i_requested_max_size; /* max_size we've requested */
299
300 /* held references to caps */
301 int i_pin_ref;
Henry C Changd3d07202011-05-11 10:29:54 +0000302 int i_rd_ref, i_rdcache_ref, i_wr_ref, i_wb_ref;
Sage Weilde576062009-10-06 11:31:07 -0700303 int i_wrbuffer_ref, i_wrbuffer_ref_head;
304 u32 i_shared_gen; /* increment each time we get FILE_SHARED */
Sage Weilcd045cb2010-11-04 11:05:05 -0700305 u32 i_rdcache_gen; /* incremented each time we get FILE_CACHE. */
Sage Weilde576062009-10-06 11:31:07 -0700306 u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */
307
308 struct list_head i_unsafe_writes; /* uncommitted sync writes */
309 struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */
310 spinlock_t i_unsafe_lock;
311
312 struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */
313 int i_snap_realm_counter; /* snap realm (if caps) */
314 struct list_head i_snap_realm_item;
315 struct list_head i_snap_flush_item;
316
317 struct work_struct i_wb_work; /* writeback work */
318 struct work_struct i_pg_inv_work; /* page invalidation work */
319
320 struct work_struct i_vmtruncate_work;
321
322 struct inode vfs_inode; /* at end */
323};
324
325static inline struct ceph_inode_info *ceph_inode(struct inode *inode)
326{
Noah Watkinsfbbccec2009-10-28 11:54:49 -0700327 return container_of(inode, struct ceph_inode_info, vfs_inode);
Sage Weilde576062009-10-06 11:31:07 -0700328}
329
Yehuda Sadehad1fee92011-01-21 16:44:03 -0800330static inline struct ceph_fs_client *ceph_inode_to_client(struct inode *inode)
331{
332 return (struct ceph_fs_client *)inode->i_sb->s_fs_info;
333}
334
335static inline struct ceph_fs_client *ceph_sb_to_client(struct super_block *sb)
336{
337 return (struct ceph_fs_client *)sb->s_fs_info;
338}
339
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700340static inline struct ceph_vino ceph_vino(struct inode *inode)
341{
342 return ceph_inode(inode)->i_vino;
343}
344
345/*
346 * ino_t is <64 bits on many architectures, blech.
347 *
Yehuda Sadehad1fee92011-01-21 16:44:03 -0800348 * i_ino (kernel inode) st_ino (userspace)
349 * i386 32 32
350 * x86_64+ino32 64 32
351 * x86_64 64 64
352 */
Amon Ott3310f752011-10-20 13:04:07 -0700353static inline u32 ceph_ino_to_ino32(__u64 vino)
Yehuda Sadehad1fee92011-01-21 16:44:03 -0800354{
Amon Ott3310f752011-10-20 13:04:07 -0700355 u32 ino = vino & 0xffffffff;
356 ino ^= vino >> 32;
Yehuda Sadehad1fee92011-01-21 16:44:03 -0800357 if (!ino)
Amon Otta661fc52012-01-23 09:25:23 -0800358 ino = 2;
Yehuda Sadehad1fee92011-01-21 16:44:03 -0800359 return ino;
360}
361
362/*
363 * kernel i_ino value
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700364 */
365static inline ino_t ceph_vino_to_ino(struct ceph_vino vino)
366{
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700367#if BITS_PER_LONG == 32
Amon Ott3310f752011-10-20 13:04:07 -0700368 return ceph_ino_to_ino32(vino.ino);
369#else
370 return (ino_t)vino.ino;
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700371#endif
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700372}
373
Yehuda Sadehad1fee92011-01-21 16:44:03 -0800374/*
375 * user-visible ino (stat, filldir)
376 */
377#if BITS_PER_LONG == 32
378static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino)
379{
380 return ino;
381}
382#else
383static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino)
384{
385 if (ceph_test_mount_opt(ceph_sb_to_client(sb), INO32))
386 ino = ceph_ino_to_ino32(ino);
387 return ino;
388}
389#endif
390
391
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700392/* for printf-style formatting */
393#define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap
394
395static inline u64 ceph_ino(struct inode *inode)
396{
397 return ceph_inode(inode)->i_vino.ino;
398}
399static inline u64 ceph_snap(struct inode *inode)
400{
401 return ceph_inode(inode)->i_vino.snap;
402}
403
404static inline int ceph_ino_compare(struct inode *inode, void *data)
405{
406 struct ceph_vino *pvino = (struct ceph_vino *)data;
407 struct ceph_inode_info *ci = ceph_inode(inode);
408 return ci->i_vino.ino == pvino->ino &&
409 ci->i_vino.snap == pvino->snap;
410}
411
412static inline struct inode *ceph_find_inode(struct super_block *sb,
413 struct ceph_vino vino)
414{
415 ino_t t = ceph_vino_to_ino(vino);
416 return ilookup5(sb, t, ceph_ino_compare, &vino);
417}
418
419
420/*
421 * Ceph inode.
422 */
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700423#define CEPH_I_NODELAY 4 /* do not delay cap release */
424#define CEPH_I_FLUSH 8 /* do not delay flush of dirty metadata */
425#define CEPH_I_NOFLUSH 16 /* do not flush dirty caps */
426
Yan, Zheng2f276c52013-03-13 19:44:32 +0800427static inline void __ceph_dir_set_complete(struct ceph_inode_info *ci,
428 int release_count)
Sage Weilde576062009-10-06 11:31:07 -0700429{
Yan, Zheng2f276c52013-03-13 19:44:32 +0800430 atomic_set(&ci->i_complete_count, release_count);
Sage Weilde576062009-10-06 11:31:07 -0700431}
432
Yan, Zheng2f276c52013-03-13 19:44:32 +0800433static inline void __ceph_dir_clear_complete(struct ceph_inode_info *ci)
Sage Weilde576062009-10-06 11:31:07 -0700434{
Yan, Zheng2f276c52013-03-13 19:44:32 +0800435 atomic_inc(&ci->i_release_count);
Sage Weilde576062009-10-06 11:31:07 -0700436}
437
Yan, Zheng2f276c52013-03-13 19:44:32 +0800438static inline bool __ceph_dir_is_complete(struct ceph_inode_info *ci)
Sage Weilde576062009-10-06 11:31:07 -0700439{
Yan, Zheng2f276c52013-03-13 19:44:32 +0800440 return atomic_read(&ci->i_complete_count) ==
441 atomic_read(&ci->i_release_count);
442}
Sage Weilde576062009-10-06 11:31:07 -0700443
Yan, Zheng2f276c52013-03-13 19:44:32 +0800444static inline void ceph_dir_clear_complete(struct inode *inode)
445{
446 __ceph_dir_clear_complete(ceph_inode(inode));
447}
448
449static inline bool ceph_dir_is_complete(struct inode *inode)
450{
451 return __ceph_dir_is_complete(ceph_inode(inode));
Sage Weilde576062009-10-06 11:31:07 -0700452}
453
454
455/* find a specific frag @f */
456extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci,
457 u32 f);
458
459/*
460 * choose fragment for value @v. copy frag content to pfrag, if leaf
461 * exists
462 */
463extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v,
464 struct ceph_inode_frag *pfrag,
465 int *found);
466
Sage Weilde576062009-10-06 11:31:07 -0700467static inline struct ceph_dentry_info *ceph_dentry(struct dentry *dentry)
468{
469 return (struct ceph_dentry_info *)dentry->d_fsdata;
470}
471
472static inline loff_t ceph_make_fpos(unsigned frag, unsigned off)
473{
474 return ((loff_t)frag << 32) | (loff_t)off;
475}
476
Sage Weilde576062009-10-06 11:31:07 -0700477/*
478 * caps helpers
479 */
480static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci)
481{
482 return !RB_EMPTY_ROOT(&ci->i_caps);
483}
484
485extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented);
486extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t);
487extern int __ceph_caps_issued_other(struct ceph_inode_info *ci,
488 struct ceph_cap *cap);
489
490static inline int ceph_caps_issued(struct ceph_inode_info *ci)
491{
492 int issued;
Sage Weilbe655592011-11-30 09:47:09 -0800493 spin_lock(&ci->i_ceph_lock);
Sage Weilde576062009-10-06 11:31:07 -0700494 issued = __ceph_caps_issued(ci, NULL);
Sage Weilbe655592011-11-30 09:47:09 -0800495 spin_unlock(&ci->i_ceph_lock);
Sage Weilde576062009-10-06 11:31:07 -0700496 return issued;
497}
498
499static inline int ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask,
500 int touch)
501{
502 int r;
Sage Weilbe655592011-11-30 09:47:09 -0800503 spin_lock(&ci->i_ceph_lock);
Sage Weilde576062009-10-06 11:31:07 -0700504 r = __ceph_caps_issued_mask(ci, mask, touch);
Sage Weilbe655592011-11-30 09:47:09 -0800505 spin_unlock(&ci->i_ceph_lock);
Sage Weilde576062009-10-06 11:31:07 -0700506 return r;
507}
508
509static inline int __ceph_caps_dirty(struct ceph_inode_info *ci)
510{
511 return ci->i_dirty_caps | ci->i_flushing_caps;
512}
Sage Weilfca65b42011-05-04 11:33:47 -0700513extern int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask);
Sage Weilde576062009-10-06 11:31:07 -0700514
515extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask);
516extern int __ceph_caps_used(struct ceph_inode_info *ci);
517
518extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci);
519
520/*
521 * wanted, by virtue of open file modes AND cap refs (buffered/cached data)
522 */
523static inline int __ceph_caps_wanted(struct ceph_inode_info *ci)
524{
525 int w = __ceph_caps_file_wanted(ci) | __ceph_caps_used(ci);
526 if (w & CEPH_CAP_FILE_BUFFER)
527 w |= CEPH_CAP_FILE_EXCL; /* we want EXCL if dirty data */
528 return w;
529}
530
531/* what the mds thinks we want */
532extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci);
533
Yehuda Sadeh37151662010-06-17 16:16:12 -0700534extern void ceph_caps_init(struct ceph_mds_client *mdsc);
535extern void ceph_caps_finalize(struct ceph_mds_client *mdsc);
536extern void ceph_adjust_min_caps(struct ceph_mds_client *mdsc, int delta);
majianpeng93faca62013-06-26 11:15:27 +0800537extern void ceph_reserve_caps(struct ceph_mds_client *mdsc,
Yehuda Sadeh37151662010-06-17 16:16:12 -0700538 struct ceph_cap_reservation *ctx, int need);
539extern int ceph_unreserve_caps(struct ceph_mds_client *mdsc,
540 struct ceph_cap_reservation *ctx);
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700541extern void ceph_reservation_status(struct ceph_fs_client *client,
Sage Weilde576062009-10-06 11:31:07 -0700542 int *total, int *avail, int *used,
Sage Weil85ccce42010-02-17 10:02:43 -0800543 int *reserved, int *min);
Sage Weilde576062009-10-06 11:31:07 -0700544
Sage Weilde576062009-10-06 11:31:07 -0700545
Sage Weilde576062009-10-06 11:31:07 -0700546
547/*
548 * we keep buffered readdir results attached to file->private_data
549 */
Sage Weil4918b6d2011-07-26 11:26:07 -0700550#define CEPH_F_SYNC 1
Sage Weil9cfa1092011-07-26 11:26:18 -0700551#define CEPH_F_ATEND 2
Sage Weil4918b6d2011-07-26 11:26:07 -0700552
Sage Weilde576062009-10-06 11:31:07 -0700553struct ceph_file_info {
Sage Weil252c6722011-07-26 11:25:27 -0700554 short fmode; /* initialized on open */
555 short flags; /* CEPH_F_* */
Sage Weilde576062009-10-06 11:31:07 -0700556
557 /* readdir: position within the dir */
558 u32 frag;
559 struct ceph_mds_request *last_readdir;
Sage Weilde576062009-10-06 11:31:07 -0700560
561 /* readdir: position within a frag */
562 unsigned offset; /* offset of last chunk, adjusted for . and .. */
563 u64 next_offset; /* offset of next chunk (last_name's + 1) */
564 char *last_name; /* last entry in previous chunk */
565 struct dentry *dentry; /* next dentry (for dcache readdir) */
Yan, Zheng2f276c52013-03-13 19:44:32 +0800566 int dir_release_count;
Sage Weilde576062009-10-06 11:31:07 -0700567
568 /* used for -o dirstat read() on directory thing */
569 char *dir_info;
570 int dir_info_len;
571};
572
573
574
575/*
Sage Weilde576062009-10-06 11:31:07 -0700576 * A "snap realm" describes a subset of the file hierarchy sharing
577 * the same set of snapshots that apply to it. The realms themselves
578 * are organized into a hierarchy, such that children inherit (some of)
579 * the snapshots of their parents.
580 *
581 * All inodes within the realm that have capabilities are linked into a
582 * per-realm list.
583 */
584struct ceph_snap_realm {
585 u64 ino;
586 atomic_t nref;
Sage Weila105f002010-02-15 14:37:55 -0800587 struct rb_node node;
588
Sage Weilde576062009-10-06 11:31:07 -0700589 u64 created, seq;
590 u64 parent_ino;
591 u64 parent_since; /* snapid when our current parent became so */
592
593 u64 *prior_parent_snaps; /* snaps inherited from any parents we */
Alex Elderaa711ee2012-07-13 20:35:11 -0500594 u32 num_prior_parent_snaps; /* had prior to parent_since */
Sage Weilde576062009-10-06 11:31:07 -0700595 u64 *snaps; /* snaps specific to this realm */
Alex Elderaa711ee2012-07-13 20:35:11 -0500596 u32 num_snaps;
Sage Weilde576062009-10-06 11:31:07 -0700597
598 struct ceph_snap_realm *parent;
599 struct list_head children; /* list of child realms */
600 struct list_head child_item;
601
602 struct list_head empty_item; /* if i have ref==0 */
603
Sage Weilae00d4f2010-09-16 16:26:51 -0700604 struct list_head dirty_item; /* if realm needs new context */
605
Sage Weilde576062009-10-06 11:31:07 -0700606 /* the current set of snaps for this realm */
607 struct ceph_snap_context *cached_context;
608
609 struct list_head inodes_with_caps;
610 spinlock_t inodes_with_caps_lock;
611};
612
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700613static inline int default_congestion_kb(void)
Sage Weilde576062009-10-06 11:31:07 -0700614{
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700615 int congestion_kb;
616
617 /*
618 * Copied from NFS
619 *
620 * congestion size, scale with available memory.
621 *
622 * 64MB: 8192k
623 * 128MB: 11585k
624 * 256MB: 16384k
625 * 512MB: 23170k
626 * 1GB: 32768k
627 * 2GB: 46340k
628 * 4GB: 65536k
629 * 8GB: 92681k
630 * 16GB: 131072k
631 *
632 * This allows larger machines to have larger/more transfers.
633 * Limit the default to 256M
634 */
635 congestion_kb = (16*int_sqrt(totalram_pages)) << (PAGE_SHIFT-10);
636 if (congestion_kb > 256*1024)
637 congestion_kb = 256*1024;
638
639 return congestion_kb;
Sage Weilde576062009-10-06 11:31:07 -0700640}
641
642
643
644/* snap.c */
645struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc,
646 u64 ino);
647extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc,
648 struct ceph_snap_realm *realm);
649extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc,
650 struct ceph_snap_realm *realm);
651extern int ceph_update_snap_trace(struct ceph_mds_client *m,
652 void *p, void *e, bool deletion);
653extern void ceph_handle_snap(struct ceph_mds_client *mdsc,
Sage Weil2600d2d2010-02-22 15:12:16 -0800654 struct ceph_mds_session *session,
Sage Weilde576062009-10-06 11:31:07 -0700655 struct ceph_msg *msg);
Sage Weilfc837c8f2010-04-13 11:41:22 -0700656extern void ceph_queue_cap_snap(struct ceph_inode_info *ci);
Sage Weilde576062009-10-06 11:31:07 -0700657extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci,
658 struct ceph_cap_snap *capsnap);
659extern void ceph_cleanup_empty_realms(struct ceph_mds_client *mdsc);
660
661/*
662 * a cap_snap is "pending" if it is still awaiting an in-progress
663 * sync write (that may/may not still update size, mtime, etc.).
664 */
665static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci)
666{
667 return !list_empty(&ci->i_cap_snaps) &&
668 list_entry(ci->i_cap_snaps.prev, struct ceph_cap_snap,
669 ci_item)->writing;
670}
671
Sage Weilde576062009-10-06 11:31:07 -0700672/* inode.c */
673extern const struct inode_operations ceph_file_iops;
674
675extern struct inode *ceph_alloc_inode(struct super_block *sb);
676extern void ceph_destroy_inode(struct inode *inode);
677
678extern struct inode *ceph_get_inode(struct super_block *sb,
679 struct ceph_vino vino);
680extern struct inode *ceph_get_snapdir(struct inode *parent);
681extern int ceph_fill_file_size(struct inode *inode, int issued,
682 u32 truncate_seq, u64 truncate_size, u64 size);
683extern void ceph_fill_file_time(struct inode *inode, int issued,
684 u64 time_warp_seq, struct timespec *ctime,
685 struct timespec *mtime, struct timespec *atime);
686extern int ceph_fill_trace(struct super_block *sb,
687 struct ceph_mds_request *req,
688 struct ceph_mds_session *session);
689extern int ceph_readdir_prepopulate(struct ceph_mds_request *req,
690 struct ceph_mds_session *session);
691
692extern int ceph_inode_holds_cap(struct inode *inode, int mask);
693
694extern int ceph_inode_set_size(struct inode *inode, loff_t size);
Yan, Zhengb415bf42013-07-02 12:40:19 +0800695extern void __ceph_do_pending_vmtruncate(struct inode *inode);
Sage Weil3c6f6b72010-02-09 15:24:44 -0800696extern void ceph_queue_vmtruncate(struct inode *inode);
697
698extern void ceph_queue_invalidate(struct inode *inode);
699extern void ceph_queue_writeback(struct inode *inode);
Sage Weilde576062009-10-06 11:31:07 -0700700
701extern int ceph_do_getattr(struct inode *inode, int mask);
Al Viro10556cb2011-06-20 19:28:19 -0400702extern int ceph_permission(struct inode *inode, int mask);
Sage Weilde576062009-10-06 11:31:07 -0700703extern int ceph_setattr(struct dentry *dentry, struct iattr *attr);
704extern int ceph_getattr(struct vfsmount *mnt, struct dentry *dentry,
705 struct kstat *stat);
706
707/* xattr.c */
708extern int ceph_setxattr(struct dentry *, const char *, const void *,
709 size_t, int);
710extern ssize_t ceph_getxattr(struct dentry *, const char *, void *, size_t);
711extern ssize_t ceph_listxattr(struct dentry *, char *, size_t);
712extern int ceph_removexattr(struct dentry *, const char *);
713extern void __ceph_build_xattrs_blob(struct ceph_inode_info *ci);
714extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci);
Alex Elder3ce6cd12012-01-23 15:49:28 -0600715extern void __init ceph_xattr_init(void);
716extern void ceph_xattr_exit(void);
Sage Weilde576062009-10-06 11:31:07 -0700717
718/* caps.c */
719extern const char *ceph_cap_string(int c);
720extern void ceph_handle_caps(struct ceph_mds_session *session,
721 struct ceph_msg *msg);
722extern int ceph_add_cap(struct inode *inode,
723 struct ceph_mds_session *session, u64 cap_id,
724 int fmode, unsigned issued, unsigned wanted,
725 unsigned cap, unsigned seq, u64 realmino, int flags,
726 struct ceph_cap_reservation *caps_reservation);
Sage Weil7c1332b2010-02-16 11:39:45 -0800727extern void __ceph_remove_cap(struct ceph_cap *cap);
Sage Weilde576062009-10-06 11:31:07 -0700728static inline void ceph_remove_cap(struct ceph_cap *cap)
729{
Sage Weilbe655592011-11-30 09:47:09 -0800730 spin_lock(&cap->ci->i_ceph_lock);
Sage Weil7c1332b2010-02-16 11:39:45 -0800731 __ceph_remove_cap(cap);
Sage Weilbe655592011-11-30 09:47:09 -0800732 spin_unlock(&cap->ci->i_ceph_lock);
Sage Weilde576062009-10-06 11:31:07 -0700733}
Yehuda Sadeh37151662010-06-17 16:16:12 -0700734extern void ceph_put_cap(struct ceph_mds_client *mdsc,
735 struct ceph_cap *cap);
Sage Weilde576062009-10-06 11:31:07 -0700736
Yan, Zhengd40ee0d2013-02-18 13:43:43 +0800737extern void __queue_cap_release(struct ceph_mds_session *session, u64 ino,
738 u64 cap_id, u32 migrate_seq, u32 issue_seq);
Sage Weilde576062009-10-06 11:31:07 -0700739extern void ceph_queue_caps_release(struct inode *inode);
Stephen Rothwellf1a3d572010-01-18 11:53:08 +1100740extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc);
Josef Bacik02c24a82011-07-16 20:44:56 -0400741extern int ceph_fsync(struct file *file, loff_t start, loff_t end,
742 int datasync);
Sage Weilde576062009-10-06 11:31:07 -0700743extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc,
744 struct ceph_mds_session *session);
Greg Farnum2bc50252010-06-30 12:44:34 -0700745extern struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci,
746 int mds);
Sage Weilde576062009-10-06 11:31:07 -0700747extern int ceph_get_cap_mds(struct inode *inode);
748extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps);
749extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had);
750extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
751 struct ceph_snap_context *snapc);
752extern void __ceph_flush_snaps(struct ceph_inode_info *ci,
Sage Weile8351242010-09-17 08:03:08 -0700753 struct ceph_mds_session **psession,
754 int again);
Sage Weilde576062009-10-06 11:31:07 -0700755extern void ceph_check_caps(struct ceph_inode_info *ci, int flags,
756 struct ceph_mds_session *session);
Sage Weilafcdaea2009-10-14 14:27:38 -0700757extern void ceph_check_delayed_caps(struct ceph_mds_client *mdsc);
758extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc);
Sage Weilde576062009-10-06 11:31:07 -0700759
760extern int ceph_encode_inode_release(void **p, struct inode *inode,
761 int mds, int drop, int unless, int force);
762extern int ceph_encode_dentry_release(void **p, struct dentry *dn,
763 int mds, int drop, int unless);
764
765extern int ceph_get_caps(struct ceph_inode_info *ci, int need, int want,
766 int *got, loff_t endoff);
767
768/* for counting open files by mode */
769static inline void __ceph_get_fmode(struct ceph_inode_info *ci, int mode)
770{
771 ci->i_nr_by_mode[mode]++;
772}
773extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode);
774
775/* addr.c */
776extern const struct address_space_operations ceph_aops;
777extern int ceph_mmap(struct file *file, struct vm_area_struct *vma);
778
779/* file.c */
780extern const struct file_operations ceph_file_fops;
781extern const struct address_space_operations ceph_aops;
Alex Elder9e0eb852013-02-06 13:11:38 -0600782
Sage Weilde576062009-10-06 11:31:07 -0700783extern int ceph_open(struct inode *inode, struct file *file);
Sage Weil5ef50c32012-07-31 11:27:36 -0700784extern int ceph_atomic_open(struct inode *dir, struct dentry *dentry,
785 struct file *file, unsigned flags, umode_t mode,
786 int *opened);
Sage Weilde576062009-10-06 11:31:07 -0700787extern int ceph_release(struct inode *inode, struct file *filp);
Sage Weilde576062009-10-06 11:31:07 -0700788
789/* dir.c */
790extern const struct file_operations ceph_dir_fops;
791extern const struct inode_operations ceph_dir_iops;
Sage Weil52dfb8a2010-08-03 10:25:30 -0700792extern const struct dentry_operations ceph_dentry_ops, ceph_snap_dentry_ops,
Sage Weilde576062009-10-06 11:31:07 -0700793 ceph_snapdir_dentry_ops;
794
795extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry);
Sage Weil468640e2011-07-26 11:28:11 -0700796extern int ceph_handle_snapdir(struct ceph_mds_request *req,
797 struct dentry *dentry, int err);
Sage Weilde576062009-10-06 11:31:07 -0700798extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req,
799 struct dentry *dentry, int err);
800
801extern void ceph_dentry_lru_add(struct dentry *dn);
802extern void ceph_dentry_lru_touch(struct dentry *dn);
803extern void ceph_dentry_lru_del(struct dentry *dn);
Sage Weil81a6cf22010-05-14 09:35:38 -0700804extern void ceph_invalidate_dentry_lease(struct dentry *dentry);
Sage Weile5f86dc2011-07-26 11:30:55 -0700805extern unsigned ceph_dentry_hash(struct inode *dir, struct dentry *dn);
Sage Weil5f21c962011-07-26 11:30:29 -0700806extern struct inode *ceph_get_dentry_parent_inode(struct dentry *dentry);
Sage Weilde576062009-10-06 11:31:07 -0700807
808/*
809 * our d_ops vary depending on whether the inode is live,
810 * snapshotted (read-only), or a virtual ".snap" directory.
811 */
812int ceph_init_dentry(struct dentry *dentry);
813
814
815/* ioctl.c */
816extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
817
818/* export.c */
819extern const struct export_operations ceph_export_ops;
820
Greg Farnum40819f62010-08-02 15:34:23 -0700821/* locks.c */
822extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl);
823extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl);
824extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num);
Jim Schutt39be95e2013-05-15 13:03:35 -0500825extern int ceph_encode_locks_to_buffer(struct inode *inode,
826 struct ceph_filelock *flocks,
827 int num_fcntl_locks,
828 int num_flock_locks);
829extern int ceph_locks_to_pagelist(struct ceph_filelock *flocks,
830 struct ceph_pagelist *pagelist,
831 int num_fcntl_locks, int num_flock_locks);
Greg Farnum40819f62010-08-02 15:34:23 -0700832extern int lock_to_ceph_filelock(struct file_lock *fl, struct ceph_filelock *c);
833
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700834/* debugfs.c */
835extern int ceph_fs_debugfs_init(struct ceph_fs_client *client);
836extern void ceph_fs_debugfs_cleanup(struct ceph_fs_client *client);
837
Sage Weilde576062009-10-06 11:31:07 -0700838#endif /* _FS_CEPH_SUPER_H */