xref: /openbmc/linux/fs/ceph/super.h (revision 930beb5a)
1 #ifndef _FS_CEPH_SUPER_H
2 #define _FS_CEPH_SUPER_H
3 
4 #include <linux/ceph/ceph_debug.h>
5 
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>
14 #include <linux/writeback.h>
15 #include <linux/slab.h>
16 
17 #include <linux/ceph/libceph.h>
18 
19 #ifdef CONFIG_CEPH_FSCACHE
20 #include <linux/fscache.h>
21 #endif
22 
23 /* f_type in struct statfs */
24 #define CEPH_SUPER_MAGIC 0x00c36400
25 
26 /* large granularity for statfs utilization stats to facilitate
27  * large volume sizes on 32-bit machines. */
28 #define CEPH_BLOCK_SHIFT   22  /* 4 MB */
29 #define CEPH_BLOCK         (1 << CEPH_BLOCK_SHIFT)
30 
31 #define CEPH_MOUNT_OPT_DIRSTAT         (1<<4) /* `cat dirname` for stats */
32 #define CEPH_MOUNT_OPT_RBYTES          (1<<5) /* dir st_bytes = rbytes */
33 #define CEPH_MOUNT_OPT_NOASYNCREADDIR  (1<<7) /* no dcache readdir */
34 #define CEPH_MOUNT_OPT_INO32           (1<<8) /* 32 bit inos */
35 #define CEPH_MOUNT_OPT_DCACHE          (1<<9) /* use dcache for readdir etc */
36 #define CEPH_MOUNT_OPT_FSCACHE         (1<<10) /* use fscache */
37 
38 #define CEPH_MOUNT_OPT_DEFAULT    (CEPH_MOUNT_OPT_RBYTES)
39 
40 #define ceph_set_mount_opt(fsc, opt) \
41 	(fsc)->mount_options->flags |= CEPH_MOUNT_OPT_##opt;
42 #define ceph_test_mount_opt(fsc, opt) \
43 	(!!((fsc)->mount_options->flags & CEPH_MOUNT_OPT_##opt))
44 
45 #define CEPH_RSIZE_DEFAULT             0           /* max read size */
46 #define CEPH_RASIZE_DEFAULT            (8192*1024) /* readahead */
47 #define CEPH_MAX_READDIR_DEFAULT        1024
48 #define CEPH_MAX_READDIR_BYTES_DEFAULT  (512*1024)
49 #define CEPH_SNAPDIRNAME_DEFAULT        ".snap"
50 
51 struct ceph_mount_options {
52 	int flags;
53 	int sb_flags;
54 
55 	int wsize;            /* max write size */
56 	int rsize;            /* max read size */
57 	int rasize;           /* max readahead */
58 	int congestion_kb;    /* max writeback in flight */
59 	int caps_wanted_delay_min, caps_wanted_delay_max;
60 	int cap_release_safety;
61 	int max_readdir;       /* max readdir result (entires) */
62 	int max_readdir_bytes; /* max readdir result (bytes) */
63 
64 	/*
65 	 * everything above this point can be memcmp'd; everything below
66 	 * is handled in compare_mount_options()
67 	 */
68 
69 	char *snapdir_name;   /* default ".snap" */
70 };
71 
72 struct ceph_fs_client {
73 	struct super_block *sb;
74 
75 	struct ceph_mount_options *mount_options;
76 	struct ceph_client *client;
77 
78 	unsigned long mount_state;
79 	int min_caps;                  /* min caps i added */
80 
81 	struct ceph_mds_client *mdsc;
82 
83 	/* writeback */
84 	mempool_t *wb_pagevec_pool;
85 	struct workqueue_struct *wb_wq;
86 	struct workqueue_struct *pg_inv_wq;
87 	struct workqueue_struct *trunc_wq;
88 	atomic_long_t writeback_count;
89 
90 	struct backing_dev_info backing_dev_info;
91 
92 #ifdef CONFIG_DEBUG_FS
93 	struct dentry *debugfs_dentry_lru, *debugfs_caps;
94 	struct dentry *debugfs_congestion_kb;
95 	struct dentry *debugfs_bdi;
96 	struct dentry *debugfs_mdsc, *debugfs_mdsmap;
97 #endif
98 
99 #ifdef CONFIG_CEPH_FSCACHE
100 	struct fscache_cookie *fscache;
101 	struct workqueue_struct *revalidate_wq;
102 #endif
103 };
104 
105 
106 /*
107  * File i/o capability.  This tracks shared state with the metadata
108  * server that allows us to cache or writeback attributes or to read
109  * and write data.  For any given inode, we should have one or more
110  * capabilities, one issued by each metadata server, and our
111  * cumulative access is the OR of all issued capabilities.
112  *
113  * Each cap is referenced by the inode's i_caps rbtree and by per-mds
114  * session capability lists.
115  */
116 struct ceph_cap {
117 	struct ceph_inode_info *ci;
118 	struct rb_node ci_node;          /* per-ci cap tree */
119 	struct ceph_mds_session *session;
120 	struct list_head session_caps;   /* per-session caplist */
121 	int mds;
122 	u64 cap_id;       /* unique cap id (mds provided) */
123 	int issued;       /* latest, from the mds */
124 	int implemented;  /* implemented superset of issued (for revocation) */
125 	int mds_wanted;
126 	u32 seq, issue_seq, mseq;
127 	u32 cap_gen;      /* active/stale cycle */
128 	unsigned long last_used;
129 	struct list_head caps_item;
130 };
131 
132 #define CHECK_CAPS_NODELAY    1  /* do not delay any further */
133 #define CHECK_CAPS_AUTHONLY   2  /* only check auth cap */
134 #define CHECK_CAPS_FLUSH      4  /* flush any dirty caps */
135 
136 /*
137  * Snapped cap state that is pending flush to mds.  When a snapshot occurs,
138  * we first complete any in-process sync writes and writeback any dirty
139  * data before flushing the snapped state (tracked here) back to the MDS.
140  */
141 struct ceph_cap_snap {
142 	atomic_t nref;
143 	struct ceph_inode_info *ci;
144 	struct list_head ci_item, flushing_item;
145 
146 	u64 follows, flush_tid;
147 	int issued, dirty;
148 	struct ceph_snap_context *context;
149 
150 	umode_t mode;
151 	kuid_t uid;
152 	kgid_t gid;
153 
154 	struct ceph_buffer *xattr_blob;
155 	u64 xattr_version;
156 
157 	u64 size;
158 	struct timespec mtime, atime, ctime;
159 	u64 time_warp_seq;
160 	int writing;   /* a sync write is still in progress */
161 	int dirty_pages;     /* dirty pages awaiting writeback */
162 };
163 
164 static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap)
165 {
166 	if (atomic_dec_and_test(&capsnap->nref)) {
167 		if (capsnap->xattr_blob)
168 			ceph_buffer_put(capsnap->xattr_blob);
169 		kfree(capsnap);
170 	}
171 }
172 
173 /*
174  * The frag tree describes how a directory is fragmented, potentially across
175  * multiple metadata servers.  It is also used to indicate points where
176  * metadata authority is delegated, and whether/where metadata is replicated.
177  *
178  * A _leaf_ frag will be present in the i_fragtree IFF there is
179  * delegation info.  That is, if mds >= 0 || ndist > 0.
180  */
181 #define CEPH_MAX_DIRFRAG_REP 4
182 
183 struct ceph_inode_frag {
184 	struct rb_node node;
185 
186 	/* fragtree state */
187 	u32 frag;
188 	int split_by;         /* i.e. 2^(split_by) children */
189 
190 	/* delegation and replication info */
191 	int mds;              /* -1 if same authority as parent */
192 	int ndist;            /* >0 if replicated */
193 	int dist[CEPH_MAX_DIRFRAG_REP];
194 };
195 
196 /*
197  * We cache inode xattrs as an encoded blob until they are first used,
198  * at which point we parse them into an rbtree.
199  */
200 struct ceph_inode_xattr {
201 	struct rb_node node;
202 
203 	const char *name;
204 	int name_len;
205 	const char *val;
206 	int val_len;
207 	int dirty;
208 
209 	int should_free_name;
210 	int should_free_val;
211 };
212 
213 /*
214  * Ceph dentry state
215  */
216 struct ceph_dentry_info {
217 	struct ceph_mds_session *lease_session;
218 	u32 lease_gen, lease_shared_gen;
219 	u32 lease_seq;
220 	unsigned long lease_renew_after, lease_renew_from;
221 	struct list_head lru;
222 	struct dentry *dentry;
223 	u64 time;
224 	u64 offset;
225 };
226 
227 struct ceph_inode_xattrs_info {
228 	/*
229 	 * (still encoded) xattr blob. we avoid the overhead of parsing
230 	 * this until someone actually calls getxattr, etc.
231 	 *
232 	 * blob->vec.iov_len == 4 implies there are no xattrs; blob ==
233 	 * NULL means we don't know.
234 	*/
235 	struct ceph_buffer *blob, *prealloc_blob;
236 
237 	struct rb_root index;
238 	bool dirty;
239 	int count;
240 	int names_size;
241 	int vals_size;
242 	u64 version, index_version;
243 };
244 
245 /*
246  * Ceph inode.
247  */
248 struct ceph_inode_info {
249 	struct ceph_vino i_vino;   /* ceph ino + snap */
250 
251 	spinlock_t i_ceph_lock;
252 
253 	u64 i_version;
254 	u32 i_time_warp_seq;
255 
256 	unsigned i_ceph_flags;
257 	atomic_t i_release_count;
258 	atomic_t i_complete_count;
259 
260 	struct ceph_dir_layout i_dir_layout;
261 	struct ceph_file_layout i_layout;
262 	char *i_symlink;
263 
264 	/* for dirs */
265 	struct timespec i_rctime;
266 	u64 i_rbytes, i_rfiles, i_rsubdirs;
267 	u64 i_files, i_subdirs;
268 	u64 i_max_offset;  /* largest readdir offset, set with complete dir */
269 
270 	struct rb_root i_fragtree;
271 	struct mutex i_fragtree_mutex;
272 
273 	struct ceph_inode_xattrs_info i_xattrs;
274 
275 	/* capabilities.  protected _both_ by i_ceph_lock and cap->session's
276 	 * s_mutex. */
277 	struct rb_root i_caps;           /* cap list */
278 	struct ceph_cap *i_auth_cap;     /* authoritative cap, if any */
279 	unsigned i_dirty_caps, i_flushing_caps;     /* mask of dirtied fields */
280 	struct list_head i_dirty_item, i_flushing_item;
281 	u64 i_cap_flush_seq;
282 	/* we need to track cap writeback on a per-cap-bit basis, to allow
283 	 * overlapping, pipelined cap flushes to the mds.  we can probably
284 	 * reduce the tid to 8 bits if we're concerned about inode size. */
285 	u16 i_cap_flush_last_tid, i_cap_flush_tid[CEPH_CAP_BITS];
286 	wait_queue_head_t i_cap_wq;      /* threads waiting on a capability */
287 	unsigned long i_hold_caps_min; /* jiffies */
288 	unsigned long i_hold_caps_max; /* jiffies */
289 	struct list_head i_cap_delay_list;  /* for delayed cap release to mds */
290 	int i_cap_exporting_mds;         /* to handle cap migration between */
291 	unsigned i_cap_exporting_mseq;   /*  mds's. */
292 	unsigned i_cap_exporting_issued;
293 	struct ceph_cap_reservation i_cap_migration_resv;
294 	struct list_head i_cap_snaps;   /* snapped state pending flush to mds */
295 	struct ceph_snap_context *i_head_snapc;  /* set if wr_buffer_head > 0 or
296 						    dirty|flushing caps */
297 	unsigned i_snap_caps;           /* cap bits for snapped files */
298 
299 	int i_nr_by_mode[CEPH_FILE_MODE_NUM];  /* open file counts */
300 
301 	struct mutex i_truncate_mutex;
302 	u32 i_truncate_seq;        /* last truncate to smaller size */
303 	u64 i_truncate_size;       /*  and the size we last truncated down to */
304 	int i_truncate_pending;    /*  still need to call vmtruncate */
305 
306 	u64 i_max_size;            /* max file size authorized by mds */
307 	u64 i_reported_size; /* (max_)size reported to or requested of mds */
308 	u64 i_wanted_max_size;     /* offset we'd like to write too */
309 	u64 i_requested_max_size;  /* max_size we've requested */
310 
311 	/* held references to caps */
312 	int i_pin_ref;
313 	int i_rd_ref, i_rdcache_ref, i_wr_ref, i_wb_ref;
314 	int i_wrbuffer_ref, i_wrbuffer_ref_head;
315 	u32 i_shared_gen;       /* increment each time we get FILE_SHARED */
316 	u32 i_rdcache_gen;      /* incremented each time we get FILE_CACHE. */
317 	u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */
318 
319 	struct list_head i_unsafe_writes; /* uncommitted sync writes */
320 	struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */
321 	spinlock_t i_unsafe_lock;
322 
323 	struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */
324 	int i_snap_realm_counter; /* snap realm (if caps) */
325 	struct list_head i_snap_realm_item;
326 	struct list_head i_snap_flush_item;
327 
328 	struct work_struct i_wb_work;  /* writeback work */
329 	struct work_struct i_pg_inv_work;  /* page invalidation work */
330 
331 	struct work_struct i_vmtruncate_work;
332 
333 #ifdef CONFIG_CEPH_FSCACHE
334 	struct fscache_cookie *fscache;
335 	u32 i_fscache_gen; /* sequence, for delayed fscache validate */
336 	struct work_struct i_revalidate_work;
337 #endif
338 
339 	struct inode vfs_inode; /* at end */
340 };
341 
342 static inline struct ceph_inode_info *ceph_inode(struct inode *inode)
343 {
344 	return container_of(inode, struct ceph_inode_info, vfs_inode);
345 }
346 
347 static inline struct ceph_fs_client *ceph_inode_to_client(struct inode *inode)
348 {
349 	return (struct ceph_fs_client *)inode->i_sb->s_fs_info;
350 }
351 
352 static inline struct ceph_fs_client *ceph_sb_to_client(struct super_block *sb)
353 {
354 	return (struct ceph_fs_client *)sb->s_fs_info;
355 }
356 
357 static inline struct ceph_vino ceph_vino(struct inode *inode)
358 {
359 	return ceph_inode(inode)->i_vino;
360 }
361 
362 /*
363  * ino_t is <64 bits on many architectures, blech.
364  *
365  *               i_ino (kernel inode)   st_ino (userspace)
366  * i386          32                     32
367  * x86_64+ino32  64                     32
368  * x86_64        64                     64
369  */
370 static inline u32 ceph_ino_to_ino32(__u64 vino)
371 {
372 	u32 ino = vino & 0xffffffff;
373 	ino ^= vino >> 32;
374 	if (!ino)
375 		ino = 2;
376 	return ino;
377 }
378 
379 /*
380  * kernel i_ino value
381  */
382 static inline ino_t ceph_vino_to_ino(struct ceph_vino vino)
383 {
384 #if BITS_PER_LONG == 32
385 	return ceph_ino_to_ino32(vino.ino);
386 #else
387 	return (ino_t)vino.ino;
388 #endif
389 }
390 
391 /*
392  * user-visible ino (stat, filldir)
393  */
394 #if BITS_PER_LONG == 32
395 static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino)
396 {
397 	return ino;
398 }
399 #else
400 static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino)
401 {
402 	if (ceph_test_mount_opt(ceph_sb_to_client(sb), INO32))
403 		ino = ceph_ino_to_ino32(ino);
404 	return ino;
405 }
406 #endif
407 
408 
409 /* for printf-style formatting */
410 #define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap
411 
412 static inline u64 ceph_ino(struct inode *inode)
413 {
414 	return ceph_inode(inode)->i_vino.ino;
415 }
416 static inline u64 ceph_snap(struct inode *inode)
417 {
418 	return ceph_inode(inode)->i_vino.snap;
419 }
420 
421 static inline int ceph_ino_compare(struct inode *inode, void *data)
422 {
423 	struct ceph_vino *pvino = (struct ceph_vino *)data;
424 	struct ceph_inode_info *ci = ceph_inode(inode);
425 	return ci->i_vino.ino == pvino->ino &&
426 		ci->i_vino.snap == pvino->snap;
427 }
428 
429 static inline struct inode *ceph_find_inode(struct super_block *sb,
430 					    struct ceph_vino vino)
431 {
432 	ino_t t = ceph_vino_to_ino(vino);
433 	return ilookup5(sb, t, ceph_ino_compare, &vino);
434 }
435 
436 
437 /*
438  * Ceph inode.
439  */
440 #define CEPH_I_NODELAY   4  /* do not delay cap release */
441 #define CEPH_I_FLUSH     8  /* do not delay flush of dirty metadata */
442 #define CEPH_I_NOFLUSH  16  /* do not flush dirty caps */
443 
444 static inline void __ceph_dir_set_complete(struct ceph_inode_info *ci,
445 					   int release_count)
446 {
447 	atomic_set(&ci->i_complete_count, release_count);
448 }
449 
450 static inline void __ceph_dir_clear_complete(struct ceph_inode_info *ci)
451 {
452 	atomic_inc(&ci->i_release_count);
453 }
454 
455 static inline bool __ceph_dir_is_complete(struct ceph_inode_info *ci)
456 {
457 	return atomic_read(&ci->i_complete_count) ==
458 		atomic_read(&ci->i_release_count);
459 }
460 
461 static inline void ceph_dir_clear_complete(struct inode *inode)
462 {
463 	__ceph_dir_clear_complete(ceph_inode(inode));
464 }
465 
466 static inline bool ceph_dir_is_complete(struct inode *inode)
467 {
468 	return __ceph_dir_is_complete(ceph_inode(inode));
469 }
470 
471 
472 /* find a specific frag @f */
473 extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci,
474 						u32 f);
475 
476 /*
477  * choose fragment for value @v.  copy frag content to pfrag, if leaf
478  * exists
479  */
480 extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v,
481 			    struct ceph_inode_frag *pfrag,
482 			    int *found);
483 
484 static inline struct ceph_dentry_info *ceph_dentry(struct dentry *dentry)
485 {
486 	return (struct ceph_dentry_info *)dentry->d_fsdata;
487 }
488 
489 static inline loff_t ceph_make_fpos(unsigned frag, unsigned off)
490 {
491 	return ((loff_t)frag << 32) | (loff_t)off;
492 }
493 
494 /*
495  * caps helpers
496  */
497 static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci)
498 {
499 	return !RB_EMPTY_ROOT(&ci->i_caps);
500 }
501 
502 extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented);
503 extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t);
504 extern int __ceph_caps_issued_other(struct ceph_inode_info *ci,
505 				    struct ceph_cap *cap);
506 
507 static inline int ceph_caps_issued(struct ceph_inode_info *ci)
508 {
509 	int issued;
510 	spin_lock(&ci->i_ceph_lock);
511 	issued = __ceph_caps_issued(ci, NULL);
512 	spin_unlock(&ci->i_ceph_lock);
513 	return issued;
514 }
515 
516 static inline int ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask,
517 					int touch)
518 {
519 	int r;
520 	spin_lock(&ci->i_ceph_lock);
521 	r = __ceph_caps_issued_mask(ci, mask, touch);
522 	spin_unlock(&ci->i_ceph_lock);
523 	return r;
524 }
525 
526 static inline int __ceph_caps_dirty(struct ceph_inode_info *ci)
527 {
528 	return ci->i_dirty_caps | ci->i_flushing_caps;
529 }
530 extern int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask);
531 
532 extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask);
533 extern int __ceph_caps_used(struct ceph_inode_info *ci);
534 
535 extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci);
536 
537 /*
538  * wanted, by virtue of open file modes AND cap refs (buffered/cached data)
539  */
540 static inline int __ceph_caps_wanted(struct ceph_inode_info *ci)
541 {
542 	int w = __ceph_caps_file_wanted(ci) | __ceph_caps_used(ci);
543 	if (w & CEPH_CAP_FILE_BUFFER)
544 		w |= CEPH_CAP_FILE_EXCL;  /* we want EXCL if dirty data */
545 	return w;
546 }
547 
548 /* what the mds thinks we want */
549 extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci);
550 
551 extern void ceph_caps_init(struct ceph_mds_client *mdsc);
552 extern void ceph_caps_finalize(struct ceph_mds_client *mdsc);
553 extern void ceph_adjust_min_caps(struct ceph_mds_client *mdsc, int delta);
554 extern void ceph_reserve_caps(struct ceph_mds_client *mdsc,
555 			     struct ceph_cap_reservation *ctx, int need);
556 extern int ceph_unreserve_caps(struct ceph_mds_client *mdsc,
557 			       struct ceph_cap_reservation *ctx);
558 extern void ceph_reservation_status(struct ceph_fs_client *client,
559 				    int *total, int *avail, int *used,
560 				    int *reserved, int *min);
561 
562 
563 
564 /*
565  * we keep buffered readdir results attached to file->private_data
566  */
567 #define CEPH_F_SYNC     1
568 #define CEPH_F_ATEND    2
569 
570 struct ceph_file_info {
571 	short fmode;     /* initialized on open */
572 	short flags;     /* CEPH_F_* */
573 
574 	/* readdir: position within the dir */
575 	u32 frag;
576 	struct ceph_mds_request *last_readdir;
577 
578 	/* readdir: position within a frag */
579 	unsigned offset;       /* offset of last chunk, adjusted for . and .. */
580 	u64 next_offset;       /* offset of next chunk (last_name's + 1) */
581 	char *last_name;       /* last entry in previous chunk */
582 	struct dentry *dentry; /* next dentry (for dcache readdir) */
583 	int dir_release_count;
584 
585 	/* used for -o dirstat read() on directory thing */
586 	char *dir_info;
587 	int dir_info_len;
588 };
589 
590 
591 
592 /*
593  * A "snap realm" describes a subset of the file hierarchy sharing
594  * the same set of snapshots that apply to it.  The realms themselves
595  * are organized into a hierarchy, such that children inherit (some of)
596  * the snapshots of their parents.
597  *
598  * All inodes within the realm that have capabilities are linked into a
599  * per-realm list.
600  */
601 struct ceph_snap_realm {
602 	u64 ino;
603 	atomic_t nref;
604 	struct rb_node node;
605 
606 	u64 created, seq;
607 	u64 parent_ino;
608 	u64 parent_since;   /* snapid when our current parent became so */
609 
610 	u64 *prior_parent_snaps;      /* snaps inherited from any parents we */
611 	u32 num_prior_parent_snaps;   /*  had prior to parent_since */
612 	u64 *snaps;                   /* snaps specific to this realm */
613 	u32 num_snaps;
614 
615 	struct ceph_snap_realm *parent;
616 	struct list_head children;       /* list of child realms */
617 	struct list_head child_item;
618 
619 	struct list_head empty_item;     /* if i have ref==0 */
620 
621 	struct list_head dirty_item;     /* if realm needs new context */
622 
623 	/* the current set of snaps for this realm */
624 	struct ceph_snap_context *cached_context;
625 
626 	struct list_head inodes_with_caps;
627 	spinlock_t inodes_with_caps_lock;
628 };
629 
630 static inline int default_congestion_kb(void)
631 {
632 	int congestion_kb;
633 
634 	/*
635 	 * Copied from NFS
636 	 *
637 	 * congestion size, scale with available memory.
638 	 *
639 	 *  64MB:    8192k
640 	 * 128MB:   11585k
641 	 * 256MB:   16384k
642 	 * 512MB:   23170k
643 	 *   1GB:   32768k
644 	 *   2GB:   46340k
645 	 *   4GB:   65536k
646 	 *   8GB:   92681k
647 	 *  16GB:  131072k
648 	 *
649 	 * This allows larger machines to have larger/more transfers.
650 	 * Limit the default to 256M
651 	 */
652 	congestion_kb = (16*int_sqrt(totalram_pages)) << (PAGE_SHIFT-10);
653 	if (congestion_kb > 256*1024)
654 		congestion_kb = 256*1024;
655 
656 	return congestion_kb;
657 }
658 
659 
660 
661 /* snap.c */
662 struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc,
663 					       u64 ino);
664 extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc,
665 				struct ceph_snap_realm *realm);
666 extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc,
667 				struct ceph_snap_realm *realm);
668 extern int ceph_update_snap_trace(struct ceph_mds_client *m,
669 				  void *p, void *e, bool deletion);
670 extern void ceph_handle_snap(struct ceph_mds_client *mdsc,
671 			     struct ceph_mds_session *session,
672 			     struct ceph_msg *msg);
673 extern void ceph_queue_cap_snap(struct ceph_inode_info *ci);
674 extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci,
675 				  struct ceph_cap_snap *capsnap);
676 extern void ceph_cleanup_empty_realms(struct ceph_mds_client *mdsc);
677 
678 /*
679  * a cap_snap is "pending" if it is still awaiting an in-progress
680  * sync write (that may/may not still update size, mtime, etc.).
681  */
682 static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci)
683 {
684 	return !list_empty(&ci->i_cap_snaps) &&
685 		list_entry(ci->i_cap_snaps.prev, struct ceph_cap_snap,
686 			   ci_item)->writing;
687 }
688 
689 /* inode.c */
690 extern const struct inode_operations ceph_file_iops;
691 
692 extern struct inode *ceph_alloc_inode(struct super_block *sb);
693 extern void ceph_destroy_inode(struct inode *inode);
694 
695 extern struct inode *ceph_get_inode(struct super_block *sb,
696 				    struct ceph_vino vino);
697 extern struct inode *ceph_get_snapdir(struct inode *parent);
698 extern int ceph_fill_file_size(struct inode *inode, int issued,
699 			       u32 truncate_seq, u64 truncate_size, u64 size);
700 extern void ceph_fill_file_time(struct inode *inode, int issued,
701 				u64 time_warp_seq, struct timespec *ctime,
702 				struct timespec *mtime, struct timespec *atime);
703 extern int ceph_fill_trace(struct super_block *sb,
704 			   struct ceph_mds_request *req,
705 			   struct ceph_mds_session *session);
706 extern int ceph_readdir_prepopulate(struct ceph_mds_request *req,
707 				    struct ceph_mds_session *session);
708 
709 extern int ceph_inode_holds_cap(struct inode *inode, int mask);
710 
711 extern int ceph_inode_set_size(struct inode *inode, loff_t size);
712 extern void __ceph_do_pending_vmtruncate(struct inode *inode);
713 extern void ceph_queue_vmtruncate(struct inode *inode);
714 
715 extern void ceph_queue_invalidate(struct inode *inode);
716 extern void ceph_queue_writeback(struct inode *inode);
717 
718 extern int ceph_do_getattr(struct inode *inode, int mask);
719 extern int ceph_permission(struct inode *inode, int mask);
720 extern int ceph_setattr(struct dentry *dentry, struct iattr *attr);
721 extern int ceph_getattr(struct vfsmount *mnt, struct dentry *dentry,
722 			struct kstat *stat);
723 
724 /* xattr.c */
725 extern int ceph_setxattr(struct dentry *, const char *, const void *,
726 			 size_t, int);
727 extern ssize_t ceph_getxattr(struct dentry *, const char *, void *, size_t);
728 extern ssize_t ceph_listxattr(struct dentry *, char *, size_t);
729 extern int ceph_removexattr(struct dentry *, const char *);
730 extern void __ceph_build_xattrs_blob(struct ceph_inode_info *ci);
731 extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci);
732 extern void __init ceph_xattr_init(void);
733 extern void ceph_xattr_exit(void);
734 
735 /* caps.c */
736 extern const char *ceph_cap_string(int c);
737 extern void ceph_handle_caps(struct ceph_mds_session *session,
738 			     struct ceph_msg *msg);
739 extern int ceph_add_cap(struct inode *inode,
740 			struct ceph_mds_session *session, u64 cap_id,
741 			int fmode, unsigned issued, unsigned wanted,
742 			unsigned cap, unsigned seq, u64 realmino, int flags,
743 			struct ceph_cap_reservation *caps_reservation);
744 extern void __ceph_remove_cap(struct ceph_cap *cap, bool queue_release);
745 extern void ceph_put_cap(struct ceph_mds_client *mdsc,
746 			 struct ceph_cap *cap);
747 
748 extern void __queue_cap_release(struct ceph_mds_session *session, u64 ino,
749 				u64 cap_id, u32 migrate_seq, u32 issue_seq);
750 extern void ceph_queue_caps_release(struct inode *inode);
751 extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc);
752 extern int ceph_fsync(struct file *file, loff_t start, loff_t end,
753 		      int datasync);
754 extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc,
755 				    struct ceph_mds_session *session);
756 extern struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci,
757 					     int mds);
758 extern int ceph_get_cap_mds(struct inode *inode);
759 extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps);
760 extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had);
761 extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
762 				       struct ceph_snap_context *snapc);
763 extern void __ceph_flush_snaps(struct ceph_inode_info *ci,
764 			       struct ceph_mds_session **psession,
765 			       int again);
766 extern void ceph_check_caps(struct ceph_inode_info *ci, int flags,
767 			    struct ceph_mds_session *session);
768 extern void ceph_check_delayed_caps(struct ceph_mds_client *mdsc);
769 extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc);
770 
771 extern int ceph_encode_inode_release(void **p, struct inode *inode,
772 				     int mds, int drop, int unless, int force);
773 extern int ceph_encode_dentry_release(void **p, struct dentry *dn,
774 				      int mds, int drop, int unless);
775 
776 extern int ceph_get_caps(struct ceph_inode_info *ci, int need, int want,
777 			 int *got, loff_t endoff);
778 
779 /* for counting open files by mode */
780 static inline void __ceph_get_fmode(struct ceph_inode_info *ci, int mode)
781 {
782 	ci->i_nr_by_mode[mode]++;
783 }
784 extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode);
785 
786 /* addr.c */
787 extern const struct address_space_operations ceph_aops;
788 extern int ceph_mmap(struct file *file, struct vm_area_struct *vma);
789 
790 /* file.c */
791 extern const struct file_operations ceph_file_fops;
792 extern const struct address_space_operations ceph_aops;
793 
794 extern int ceph_open(struct inode *inode, struct file *file);
795 extern int ceph_atomic_open(struct inode *dir, struct dentry *dentry,
796 			    struct file *file, unsigned flags, umode_t mode,
797 			    int *opened);
798 extern int ceph_release(struct inode *inode, struct file *filp);
799 
800 /* dir.c */
801 extern const struct file_operations ceph_dir_fops;
802 extern const struct inode_operations ceph_dir_iops;
803 extern const struct dentry_operations ceph_dentry_ops, ceph_snap_dentry_ops,
804 	ceph_snapdir_dentry_ops;
805 
806 extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry);
807 extern int ceph_handle_snapdir(struct ceph_mds_request *req,
808 			       struct dentry *dentry, int err);
809 extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req,
810 					 struct dentry *dentry, int err);
811 
812 extern void ceph_dentry_lru_add(struct dentry *dn);
813 extern void ceph_dentry_lru_touch(struct dentry *dn);
814 extern void ceph_dentry_lru_del(struct dentry *dn);
815 extern void ceph_invalidate_dentry_lease(struct dentry *dentry);
816 extern unsigned ceph_dentry_hash(struct inode *dir, struct dentry *dn);
817 extern struct inode *ceph_get_dentry_parent_inode(struct dentry *dentry);
818 
819 /*
820  * our d_ops vary depending on whether the inode is live,
821  * snapshotted (read-only), or a virtual ".snap" directory.
822  */
823 int ceph_init_dentry(struct dentry *dentry);
824 
825 
826 /* ioctl.c */
827 extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
828 
829 /* export.c */
830 extern const struct export_operations ceph_export_ops;
831 
832 /* locks.c */
833 extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl);
834 extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl);
835 extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num);
836 extern int ceph_encode_locks_to_buffer(struct inode *inode,
837 				       struct ceph_filelock *flocks,
838 				       int num_fcntl_locks,
839 				       int num_flock_locks);
840 extern int ceph_locks_to_pagelist(struct ceph_filelock *flocks,
841 				  struct ceph_pagelist *pagelist,
842 				  int num_fcntl_locks, int num_flock_locks);
843 extern int lock_to_ceph_filelock(struct file_lock *fl, struct ceph_filelock *c);
844 
845 /* debugfs.c */
846 extern int ceph_fs_debugfs_init(struct ceph_fs_client *client);
847 extern void ceph_fs_debugfs_cleanup(struct ceph_fs_client *client);
848 
849 #endif /* _FS_CEPH_SUPER_H */
850