Linux Audio

Check our new training course

Loading...
v4.6
 
  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#include <linux/posix_acl.h>
 17
 18#include <linux/ceph/libceph.h>
 19
 20#ifdef CONFIG_CEPH_FSCACHE
 21#include <linux/fscache.h>
 22#endif
 23
 24/* f_type in struct statfs */
 25#define CEPH_SUPER_MAGIC 0x00c36400
 26
 27/* large granularity for statfs utilization stats to facilitate
 28 * large volume sizes on 32-bit machines. */
 29#define CEPH_BLOCK_SHIFT   22  /* 4 MB */
 30#define CEPH_BLOCK         (1 << CEPH_BLOCK_SHIFT)
 
 31
 
 32#define CEPH_MOUNT_OPT_DIRSTAT         (1<<4) /* `cat dirname` for stats */
 33#define CEPH_MOUNT_OPT_RBYTES          (1<<5) /* dir st_bytes = rbytes */
 34#define CEPH_MOUNT_OPT_NOASYNCREADDIR  (1<<7) /* no dcache readdir */
 35#define CEPH_MOUNT_OPT_INO32           (1<<8) /* 32 bit inos */
 36#define CEPH_MOUNT_OPT_DCACHE          (1<<9) /* use dcache for readdir etc */
 37#define CEPH_MOUNT_OPT_FSCACHE         (1<<10) /* use fscache */
 38#define CEPH_MOUNT_OPT_NOPOOLPERM      (1<<11) /* no pool permission check */
 39
 40#define CEPH_MOUNT_OPT_DEFAULT    CEPH_MOUNT_OPT_DCACHE
 
 
 
 
 
 
 
 
 
 41
 42#define ceph_set_mount_opt(fsc, opt) \
 43	(fsc)->mount_options->flags |= CEPH_MOUNT_OPT_##opt;
 
 
 44#define ceph_test_mount_opt(fsc, opt) \
 45	(!!((fsc)->mount_options->flags & CEPH_MOUNT_OPT_##opt))
 46
 47#define CEPH_RSIZE_DEFAULT             0           /* max read size */
 48#define CEPH_RASIZE_DEFAULT            (8192*1024) /* readahead */
 
 
 
 
 49#define CEPH_MAX_READDIR_DEFAULT        1024
 50#define CEPH_MAX_READDIR_BYTES_DEFAULT  (512*1024)
 51#define CEPH_SNAPDIRNAME_DEFAULT        ".snap"
 52
 
 
 
 
 
 
 
 
 
 53struct ceph_mount_options {
 54	int flags;
 55	int sb_flags;
 56
 57	int wsize;            /* max write size */
 58	int rsize;            /* max read size */
 59	int rasize;           /* max readahead */
 60	int congestion_kb;    /* max writeback in flight */
 61	int caps_wanted_delay_min, caps_wanted_delay_max;
 62	int cap_release_safety;
 63	int max_readdir;       /* max readdir result (entires) */
 64	int max_readdir_bytes; /* max readdir result (bytes) */
 
 
 65
 66	/*
 67	 * everything above this point can be memcmp'd; everything below
 68	 * is handled in compare_mount_options()
 69	 */
 70
 71	char *snapdir_name;   /* default ".snap" */
 
 
 
 
 
 72};
 73
 
 
 
 
 
 
 
 
 
 
 
 
 
 74struct ceph_fs_client {
 75	struct super_block *sb;
 76
 
 
 77	struct ceph_mount_options *mount_options;
 78	struct ceph_client *client;
 79
 80	unsigned long mount_state;
 81	int min_caps;                  /* min caps i added */
 
 
 
 
 
 
 82
 83	struct ceph_mds_client *mdsc;
 84
 85	/* writeback */
 86	mempool_t *wb_pagevec_pool;
 87	struct workqueue_struct *wb_wq;
 88	struct workqueue_struct *pg_inv_wq;
 89	struct workqueue_struct *trunc_wq;
 90	atomic_long_t writeback_count;
 
 
 
 
 91
 92	struct backing_dev_info backing_dev_info;
 
 93
 94#ifdef CONFIG_DEBUG_FS
 95	struct dentry *debugfs_dentry_lru, *debugfs_caps;
 96	struct dentry *debugfs_congestion_kb;
 97	struct dentry *debugfs_bdi;
 98	struct dentry *debugfs_mdsc, *debugfs_mdsmap;
 
 99	struct dentry *debugfs_mds_sessions;
 
100#endif
101
102#ifdef CONFIG_CEPH_FSCACHE
103	struct fscache_cookie *fscache;
104	struct workqueue_struct *revalidate_wq;
 
 
105#endif
106};
107
108
109/*
110 * File i/o capability.  This tracks shared state with the metadata
111 * server that allows us to cache or writeback attributes or to read
112 * and write data.  For any given inode, we should have one or more
113 * capabilities, one issued by each metadata server, and our
114 * cumulative access is the OR of all issued capabilities.
115 *
116 * Each cap is referenced by the inode's i_caps rbtree and by per-mds
117 * session capability lists.
118 */
119struct ceph_cap {
120	struct ceph_inode_info *ci;
121	struct rb_node ci_node;          /* per-ci cap tree */
122	struct ceph_mds_session *session;
123	struct list_head session_caps;   /* per-session caplist */
124	u64 cap_id;       /* unique cap id (mds provided) */
125	union {
126		/* in-use caps */
127		struct {
128			int issued;       /* latest, from the mds */
129			int implemented;  /* implemented superset of
130					     issued (for revocation) */
131			int mds, mds_wanted;
 
132		};
133		/* caps to release */
134		struct {
135			u64 cap_ino;
136			int queue_release;
137		};
138	};
139	u32 seq, issue_seq, mseq;
140	u32 cap_gen;      /* active/stale cycle */
141	unsigned long last_used;
142	struct list_head caps_item;
143};
144
145#define CHECK_CAPS_NODELAY    1  /* do not delay any further */
146#define CHECK_CAPS_AUTHONLY   2  /* only check auth cap */
147#define CHECK_CAPS_FLUSH      4  /* flush any dirty caps */
 
 
 
 
 
 
 
 
 
 
148
149/*
150 * Snapped cap state that is pending flush to mds.  When a snapshot occurs,
151 * we first complete any in-process sync writes and writeback any dirty
152 * data before flushing the snapped state (tracked here) back to the MDS.
153 */
154struct ceph_cap_snap {
155	atomic_t nref;
156	struct ceph_inode_info *ci;
157	struct list_head ci_item, flushing_item;
 
158
159	u64 follows, flush_tid;
160	int issued, dirty;
161	struct ceph_snap_context *context;
162
163	umode_t mode;
164	kuid_t uid;
165	kgid_t gid;
166
167	struct ceph_buffer *xattr_blob;
168	u64 xattr_version;
169
170	u64 size;
171	struct timespec mtime, atime, ctime;
 
172	u64 time_warp_seq;
 
 
173	int writing;   /* a sync write is still in progress */
174	int dirty_pages;     /* dirty pages awaiting writeback */
175	bool inline_data;
176	bool need_flush;
177};
178
179static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap)
180{
181	if (atomic_dec_and_test(&capsnap->nref)) {
182		if (capsnap->xattr_blob)
183			ceph_buffer_put(capsnap->xattr_blob);
184		kfree(capsnap);
185	}
186}
187
188struct ceph_cap_flush {
189	u64 tid;
190	int caps;
191	struct rb_node g_node; // global
192	union {
193		struct rb_node i_node; // inode
194		struct list_head list;
195	};
196};
197
198/*
199 * The frag tree describes how a directory is fragmented, potentially across
200 * multiple metadata servers.  It is also used to indicate points where
201 * metadata authority is delegated, and whether/where metadata is replicated.
202 *
203 * A _leaf_ frag will be present in the i_fragtree IFF there is
204 * delegation info.  That is, if mds >= 0 || ndist > 0.
205 */
206#define CEPH_MAX_DIRFRAG_REP 4
207
208struct ceph_inode_frag {
209	struct rb_node node;
210
211	/* fragtree state */
212	u32 frag;
213	int split_by;         /* i.e. 2^(split_by) children */
214
215	/* delegation and replication info */
216	int mds;              /* -1 if same authority as parent */
217	int ndist;            /* >0 if replicated */
218	int dist[CEPH_MAX_DIRFRAG_REP];
219};
220
221/*
222 * We cache inode xattrs as an encoded blob until they are first used,
223 * at which point we parse them into an rbtree.
224 */
225struct ceph_inode_xattr {
226	struct rb_node node;
227
228	const char *name;
229	int name_len;
230	const char *val;
231	int val_len;
232	int dirty;
233
234	int should_free_name;
235	int should_free_val;
236};
237
238/*
239 * Ceph dentry state
240 */
241struct ceph_dentry_info {
 
242	struct ceph_mds_session *lease_session;
243	u32 lease_gen, lease_shared_gen;
 
 
 
 
244	u32 lease_seq;
245	unsigned long lease_renew_after, lease_renew_from;
246	struct list_head lru;
247	struct dentry *dentry;
248	u64 time;
249	u64 offset;
250};
251
 
 
 
 
 
 
 
 
 
252struct ceph_inode_xattrs_info {
253	/*
254	 * (still encoded) xattr blob. we avoid the overhead of parsing
255	 * this until someone actually calls getxattr, etc.
256	 *
257	 * blob->vec.iov_len == 4 implies there are no xattrs; blob ==
258	 * NULL means we don't know.
259	*/
260	struct ceph_buffer *blob, *prealloc_blob;
261
262	struct rb_root index;
263	bool dirty;
264	int count;
265	int names_size;
266	int vals_size;
267	u64 version, index_version;
268};
269
270/*
271 * Ceph inode.
272 */
273struct ceph_inode_info {
 
274	struct ceph_vino i_vino;   /* ceph ino + snap */
275
276	spinlock_t i_ceph_lock;
277
278	u64 i_version;
279	u64 i_inline_version;
280	u32 i_time_warp_seq;
281
282	unsigned i_ceph_flags;
283	atomic64_t i_release_count;
284	atomic64_t i_ordered_count;
285	atomic64_t i_complete_seq[2];
286
287	struct ceph_dir_layout i_dir_layout;
288	struct ceph_file_layout i_layout;
289	size_t i_pool_ns_len;
290	char *i_symlink;
291
292	/* for dirs */
293	struct timespec i_rctime;
294	u64 i_rbytes, i_rfiles, i_rsubdirs;
295	u64 i_files, i_subdirs;
296
 
 
 
 
 
297	struct rb_root i_fragtree;
 
298	struct mutex i_fragtree_mutex;
299
300	struct ceph_inode_xattrs_info i_xattrs;
301
302	/* capabilities.  protected _both_ by i_ceph_lock and cap->session's
303	 * s_mutex. */
304	struct rb_root i_caps;           /* cap list */
305	struct ceph_cap *i_auth_cap;     /* authoritative cap, if any */
306	unsigned i_dirty_caps, i_flushing_caps;     /* mask of dirtied fields */
307	struct list_head i_dirty_item, i_flushing_item;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
308	/* we need to track cap writeback on a per-cap-bit basis, to allow
309	 * overlapping, pipelined cap flushes to the mds.  we can probably
310	 * reduce the tid to 8 bits if we're concerned about inode size. */
311	struct ceph_cap_flush *i_prealloc_cap_flush;
312	struct rb_root i_cap_flush_tree;
313	wait_queue_head_t i_cap_wq;      /* threads waiting on a capability */
314	unsigned long i_hold_caps_min; /* jiffies */
315	unsigned long i_hold_caps_max; /* jiffies */
316	struct list_head i_cap_delay_list;  /* for delayed cap release to mds */
317	struct ceph_cap_reservation i_cap_migration_resv;
318	struct list_head i_cap_snaps;   /* snapped state pending flush to mds */
319	struct ceph_snap_context *i_head_snapc;  /* set if wr_buffer_head > 0 or
320						    dirty|flushing caps */
321	unsigned i_snap_caps;           /* cap bits for snapped files */
322
323	int i_nr_by_mode[CEPH_FILE_MODE_NUM];  /* open file counts */
 
 
324
325	struct mutex i_truncate_mutex;
326	u32 i_truncate_seq;        /* last truncate to smaller size */
327	u64 i_truncate_size;       /*  and the size we last truncated down to */
328	int i_truncate_pending;    /*  still need to call vmtruncate */
 
 
 
 
 
329
330	u64 i_max_size;            /* max file size authorized by mds */
331	u64 i_reported_size; /* (max_)size reported to or requested of mds */
332	u64 i_wanted_max_size;     /* offset we'd like to write too */
333	u64 i_requested_max_size;  /* max_size we've requested */
334
335	/* held references to caps */
336	int i_pin_ref;
337	int i_rd_ref, i_rdcache_ref, i_wr_ref, i_wb_ref;
338	int i_wrbuffer_ref, i_wrbuffer_ref_head;
339	u32 i_shared_gen;       /* increment each time we get FILE_SHARED */
 
340	u32 i_rdcache_gen;      /* incremented each time we get FILE_CACHE. */
341	u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */
342
343	struct list_head i_unsafe_writes; /* uncommitted sync writes */
344	struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */
345	struct list_head i_unsafe_iops;   /* uncommitted mds inode ops */
346	spinlock_t i_unsafe_lock;
347
348	struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */
349	int i_snap_realm_counter; /* snap realm (if caps) */
 
 
350	struct list_head i_snap_realm_item;
351	struct list_head i_snap_flush_item;
 
 
352
353	struct work_struct i_wb_work;  /* writeback work */
354	struct work_struct i_pg_inv_work;  /* page invalidation work */
355
356	struct work_struct i_vmtruncate_work;
357
358#ifdef CONFIG_CEPH_FSCACHE
359	struct fscache_cookie *fscache;
360	u32 i_fscache_gen; /* sequence, for delayed fscache validate */
361	struct work_struct i_revalidate_work;
 
362#endif
363	struct inode vfs_inode; /* at end */
364};
365
366static inline struct ceph_inode_info *ceph_inode(struct inode *inode)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
367{
368	return container_of(inode, struct ceph_inode_info, vfs_inode);
369}
370
371static inline struct ceph_fs_client *ceph_inode_to_client(struct inode *inode)
 
372{
373	return (struct ceph_fs_client *)inode->i_sb->s_fs_info;
374}
375
376static inline struct ceph_fs_client *ceph_sb_to_client(struct super_block *sb)
 
377{
378	return (struct ceph_fs_client *)sb->s_fs_info;
379}
380
381static inline struct ceph_vino ceph_vino(struct inode *inode)
 
 
 
 
 
 
 
 
 
 
 
 
 
382{
383	return ceph_inode(inode)->i_vino;
384}
385
386/*
387 * ino_t is <64 bits on many architectures, blech.
388 *
389 *               i_ino (kernel inode)   st_ino (userspace)
390 * i386          32                     32
391 * x86_64+ino32  64                     32
392 * x86_64        64                     64
393 */
394static inline u32 ceph_ino_to_ino32(__u64 vino)
395{
396	u32 ino = vino & 0xffffffff;
397	ino ^= vino >> 32;
398	if (!ino)
399		ino = 2;
400	return ino;
401}
402
403/*
404 * kernel i_ino value
 
 
 
405 */
406static inline ino_t ceph_vino_to_ino(struct ceph_vino vino)
407{
408#if BITS_PER_LONG == 32
409	return ceph_ino_to_ino32(vino.ino);
410#else
411	return (ino_t)vino.ino;
412#endif
413}
414
415/*
416 * user-visible ino (stat, filldir)
417 */
418#if BITS_PER_LONG == 32
419static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino)
420{
421	return ino;
422}
423#else
424static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino)
425{
426	if (ceph_test_mount_opt(ceph_sb_to_client(sb), INO32))
427		ino = ceph_ino_to_ino32(ino);
428	return ino;
429}
430#endif
431
432
433/* for printf-style formatting */
434#define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap
435
436static inline u64 ceph_ino(struct inode *inode)
437{
438	return ceph_inode(inode)->i_vino.ino;
439}
 
440static inline u64 ceph_snap(struct inode *inode)
441{
442	return ceph_inode(inode)->i_vino.snap;
443}
444
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
445static inline int ceph_ino_compare(struct inode *inode, void *data)
446{
447	struct ceph_vino *pvino = (struct ceph_vino *)data;
448	struct ceph_inode_info *ci = ceph_inode(inode);
449	return ci->i_vino.ino == pvino->ino &&
450		ci->i_vino.snap == pvino->snap;
451}
452
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
453static inline struct inode *ceph_find_inode(struct super_block *sb,
454					    struct ceph_vino vino)
455{
456	ino_t t = ceph_vino_to_ino(vino);
457	return ilookup5(sb, t, ceph_ino_compare, &vino);
 
 
 
 
 
 
 
458}
459
460
461/*
462 * Ceph inode.
463 */
464#define CEPH_I_DIR_ORDERED	(1 << 0)  /* dentries in dir are ordered */
465#define CEPH_I_NODELAY		(1 << 1)  /* do not delay cap release */
466#define CEPH_I_FLUSH		(1 << 2)  /* do not delay flush of dirty metadata */
467#define CEPH_I_NOFLUSH		(1 << 3)  /* do not flush dirty caps */
468#define CEPH_I_POOL_PERM	(1 << 4)  /* pool rd/wr bits are valid */
469#define CEPH_I_POOL_RD		(1 << 5)  /* can read from pool */
470#define CEPH_I_POOL_WR		(1 << 6)  /* can write to pool */
471#define CEPH_I_SEC_INITED	(1 << 7)  /* security initialized */
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
472
473static inline void __ceph_dir_set_complete(struct ceph_inode_info *ci,
474					   long long release_count,
475					   long long ordered_count)
476{
477	smp_mb__before_atomic();
 
 
 
 
 
478	atomic64_set(&ci->i_complete_seq[0], release_count);
479	atomic64_set(&ci->i_complete_seq[1], ordered_count);
480}
481
482static inline void __ceph_dir_clear_complete(struct ceph_inode_info *ci)
483{
484	atomic64_inc(&ci->i_release_count);
485}
486
487static inline void __ceph_dir_clear_ordered(struct ceph_inode_info *ci)
488{
489	atomic64_inc(&ci->i_ordered_count);
490}
491
492static inline bool __ceph_dir_is_complete(struct ceph_inode_info *ci)
493{
494	return atomic64_read(&ci->i_complete_seq[0]) ==
495		atomic64_read(&ci->i_release_count);
496}
497
498static inline bool __ceph_dir_is_complete_ordered(struct ceph_inode_info *ci)
499{
500	return  atomic64_read(&ci->i_complete_seq[0]) ==
501		atomic64_read(&ci->i_release_count) &&
502		atomic64_read(&ci->i_complete_seq[1]) ==
503		atomic64_read(&ci->i_ordered_count);
504}
505
506static inline void ceph_dir_clear_complete(struct inode *inode)
507{
508	__ceph_dir_clear_complete(ceph_inode(inode));
509}
510
511static inline void ceph_dir_clear_ordered(struct inode *inode)
512{
513	__ceph_dir_clear_ordered(ceph_inode(inode));
514}
515
516static inline bool ceph_dir_is_complete_ordered(struct inode *inode)
517{
518	bool ret = __ceph_dir_is_complete_ordered(ceph_inode(inode));
519	smp_rmb();
520	return ret;
521}
522
523/* find a specific frag @f */
524extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci,
525						u32 f);
526
527/*
528 * choose fragment for value @v.  copy frag content to pfrag, if leaf
529 * exists
530 */
531extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v,
532			    struct ceph_inode_frag *pfrag,
533			    int *found);
534
535static inline struct ceph_dentry_info *ceph_dentry(struct dentry *dentry)
536{
537	return (struct ceph_dentry_info *)dentry->d_fsdata;
538}
539
540static inline loff_t ceph_make_fpos(unsigned frag, unsigned off)
541{
542	return ((loff_t)frag << 32) | (loff_t)off;
543}
544
545/*
546 * caps helpers
547 */
548static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci)
549{
550	return !RB_EMPTY_ROOT(&ci->i_caps);
551}
552
553extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented);
554extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t);
 
 
555extern int __ceph_caps_issued_other(struct ceph_inode_info *ci,
556				    struct ceph_cap *cap);
557
558static inline int ceph_caps_issued(struct ceph_inode_info *ci)
559{
560	int issued;
561	spin_lock(&ci->i_ceph_lock);
562	issued = __ceph_caps_issued(ci, NULL);
563	spin_unlock(&ci->i_ceph_lock);
564	return issued;
565}
566
567static inline int ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask,
568					int touch)
569{
570	int r;
571	spin_lock(&ci->i_ceph_lock);
572	r = __ceph_caps_issued_mask(ci, mask, touch);
573	spin_unlock(&ci->i_ceph_lock);
574	return r;
575}
576
577static inline int __ceph_caps_dirty(struct ceph_inode_info *ci)
578{
579	return ci->i_dirty_caps | ci->i_flushing_caps;
580}
581extern struct ceph_cap_flush *ceph_alloc_cap_flush(void);
582extern void ceph_free_cap_flush(struct ceph_cap_flush *cf);
583extern int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask,
584				  struct ceph_cap_flush **pcf);
585
586extern int __ceph_caps_revoking_other(struct ceph_inode_info *ci,
587				      struct ceph_cap *ocap, int mask);
588extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask);
589extern int __ceph_caps_used(struct ceph_inode_info *ci);
590
591extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci);
592
593/*
594 * wanted, by virtue of open file modes AND cap refs (buffered/cached data)
595 */
596static inline int __ceph_caps_wanted(struct ceph_inode_info *ci)
597{
598	int w = __ceph_caps_file_wanted(ci) | __ceph_caps_used(ci);
599	if (w & CEPH_CAP_FILE_BUFFER)
600		w |= CEPH_CAP_FILE_EXCL;  /* we want EXCL if dirty data */
601	return w;
602}
 
 
603
604/* what the mds thinks we want */
605extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci);
606
607extern void ceph_caps_init(struct ceph_mds_client *mdsc);
608extern void ceph_caps_finalize(struct ceph_mds_client *mdsc);
609extern void ceph_adjust_min_caps(struct ceph_mds_client *mdsc, int delta);
610extern void ceph_reserve_caps(struct ceph_mds_client *mdsc,
 
611			     struct ceph_cap_reservation *ctx, int need);
612extern int ceph_unreserve_caps(struct ceph_mds_client *mdsc,
613			       struct ceph_cap_reservation *ctx);
614extern void ceph_reservation_status(struct ceph_fs_client *client,
615				    int *total, int *avail, int *used,
616				    int *reserved, int *min);
 
 
617
618
619
620/*
621 * we keep buffered readdir results attached to file->private_data
622 */
623#define CEPH_F_SYNC     1
624#define CEPH_F_ATEND    2
625
626struct ceph_file_info {
627	short fmode;     /* initialized on open */
628	short flags;     /* CEPH_F_* */
629
 
 
 
 
 
 
 
 
 
630	/* readdir: position within the dir */
631	u32 frag;
632	struct ceph_mds_request *last_readdir;
633
634	/* readdir: position within a frag */
635	unsigned offset;       /* offset of last chunk, adjusted for . and .. */
636	unsigned next_offset;  /* offset of next chunk (last_name's + 1) */
637	char *last_name;       /* last entry in previous chunk */
638	long long dir_release_count;
639	long long dir_ordered_count;
640	int readdir_cache_idx;
641
642	/* used for -o dirstat read() on directory thing */
643	char *dir_info;
644	int dir_info_len;
645};
646
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
647struct ceph_readdir_cache_control {
648	struct page  *page;
649	struct dentry **dentries;
650	int index;
651};
652
653/*
654 * A "snap realm" describes a subset of the file hierarchy sharing
655 * the same set of snapshots that apply to it.  The realms themselves
656 * are organized into a hierarchy, such that children inherit (some of)
657 * the snapshots of their parents.
658 *
659 * All inodes within the realm that have capabilities are linked into a
660 * per-realm list.
661 */
662struct ceph_snap_realm {
663	u64 ino;
 
664	atomic_t nref;
665	struct rb_node node;
666
667	u64 created, seq;
668	u64 parent_ino;
669	u64 parent_since;   /* snapid when our current parent became so */
670
671	u64 *prior_parent_snaps;      /* snaps inherited from any parents we */
672	u32 num_prior_parent_snaps;   /*  had prior to parent_since */
673	u64 *snaps;                   /* snaps specific to this realm */
674	u32 num_snaps;
675
676	struct ceph_snap_realm *parent;
677	struct list_head children;       /* list of child realms */
678	struct list_head child_item;
679
680	struct list_head empty_item;     /* if i have ref==0 */
681
682	struct list_head dirty_item;     /* if realm needs new context */
683
 
 
684	/* the current set of snaps for this realm */
685	struct ceph_snap_context *cached_context;
686
687	struct list_head inodes_with_caps;
688	spinlock_t inodes_with_caps_lock;
689};
690
691static inline int default_congestion_kb(void)
692{
693	int congestion_kb;
694
695	/*
696	 * Copied from NFS
697	 *
698	 * congestion size, scale with available memory.
699	 *
700	 *  64MB:    8192k
701	 * 128MB:   11585k
702	 * 256MB:   16384k
703	 * 512MB:   23170k
704	 *   1GB:   32768k
705	 *   2GB:   46340k
706	 *   4GB:   65536k
707	 *   8GB:   92681k
708	 *  16GB:  131072k
709	 *
710	 * This allows larger machines to have larger/more transfers.
711	 * Limit the default to 256M
712	 */
713	congestion_kb = (16*int_sqrt(totalram_pages)) << (PAGE_SHIFT-10);
714	if (congestion_kb > 256*1024)
715		congestion_kb = 256*1024;
716
717	return congestion_kb;
718}
719
720
721
 
722/* snap.c */
723struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc,
724					       u64 ino);
725extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc,
726				struct ceph_snap_realm *realm);
727extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc,
728				struct ceph_snap_realm *realm);
729extern int ceph_update_snap_trace(struct ceph_mds_client *m,
730				  void *p, void *e, bool deletion,
731				  struct ceph_snap_realm **realm_ret);
 
732extern void ceph_handle_snap(struct ceph_mds_client *mdsc,
733			     struct ceph_mds_session *session,
734			     struct ceph_msg *msg);
735extern void ceph_queue_cap_snap(struct ceph_inode_info *ci);
736extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci,
737				  struct ceph_cap_snap *capsnap);
738extern void ceph_cleanup_empty_realms(struct ceph_mds_client *mdsc);
 
 
 
 
 
 
 
 
 
739
740/*
741 * a cap_snap is "pending" if it is still awaiting an in-progress
742 * sync write (that may/may not still update size, mtime, etc.).
743 */
744static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci)
745{
746	return !list_empty(&ci->i_cap_snaps) &&
747	       list_last_entry(&ci->i_cap_snaps, struct ceph_cap_snap,
748			       ci_item)->writing;
749}
750
751/* inode.c */
 
 
 
 
752extern const struct inode_operations ceph_file_iops;
753
754extern struct inode *ceph_alloc_inode(struct super_block *sb);
755extern void ceph_destroy_inode(struct inode *inode);
756extern int ceph_drop_inode(struct inode *inode);
 
 
 
 
 
757
758extern struct inode *ceph_get_inode(struct super_block *sb,
759				    struct ceph_vino vino);
 
760extern struct inode *ceph_get_snapdir(struct inode *parent);
761extern int ceph_fill_file_size(struct inode *inode, int issued,
762			       u32 truncate_seq, u64 truncate_size, u64 size);
763extern void ceph_fill_file_time(struct inode *inode, int issued,
764				u64 time_warp_seq, struct timespec *ctime,
765				struct timespec *mtime, struct timespec *atime);
 
 
 
 
 
 
766extern int ceph_fill_trace(struct super_block *sb,
767			   struct ceph_mds_request *req,
768			   struct ceph_mds_session *session);
769extern int ceph_readdir_prepopulate(struct ceph_mds_request *req,
770				    struct ceph_mds_session *session);
771
772extern int ceph_inode_holds_cap(struct inode *inode, int mask);
773
774extern int ceph_inode_set_size(struct inode *inode, loff_t size);
775extern void __ceph_do_pending_vmtruncate(struct inode *inode);
776extern void ceph_queue_vmtruncate(struct inode *inode);
777
778extern void ceph_queue_invalidate(struct inode *inode);
779extern void ceph_queue_writeback(struct inode *inode);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
780
 
 
 
 
 
 
 
 
 
 
 
781extern int __ceph_do_getattr(struct inode *inode, struct page *locked_page,
782			     int mask, bool force);
783static inline int ceph_do_getattr(struct inode *inode, int mask, bool force)
784{
785	return __ceph_do_getattr(inode, NULL, mask, force);
786}
787extern int ceph_permission(struct inode *inode, int mask);
788extern int ceph_setattr(struct dentry *dentry, struct iattr *attr);
789extern int ceph_getattr(struct vfsmount *mnt, struct dentry *dentry,
790			struct kstat *stat);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
791
792/* xattr.c */
793extern int ceph_setxattr(struct dentry *, const char *, const void *,
794			 size_t, int);
795int __ceph_setxattr(struct dentry *, const char *, const void *, size_t, int);
796ssize_t __ceph_getxattr(struct inode *, const char *, void *, size_t);
797int __ceph_removexattr(struct dentry *, const char *);
798extern ssize_t ceph_getxattr(struct dentry *, const char *, void *, size_t);
799extern ssize_t ceph_listxattr(struct dentry *, char *, size_t);
800extern int ceph_removexattr(struct dentry *, const char *);
801extern void __ceph_build_xattrs_blob(struct ceph_inode_info *ci);
802extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci);
803extern void __init ceph_xattr_init(void);
804extern void ceph_xattr_exit(void);
805extern const struct xattr_handler *ceph_xattr_handlers[];
 
 
 
 
 
 
 
 
 
 
 
 
 
806
807#ifdef CONFIG_SECURITY
808extern bool ceph_security_xattr_deadlock(struct inode *in);
809extern bool ceph_security_xattr_wanted(struct inode *in);
810#else
811static inline bool ceph_security_xattr_deadlock(struct inode *in)
812{
813	return false;
814}
815static inline bool ceph_security_xattr_wanted(struct inode *in)
816{
817	return false;
818}
819#endif
820
821/* acl.c */
822struct ceph_acls_info {
823	void *default_acl;
824	void *acl;
825	struct ceph_pagelist *pagelist;
826};
 
 
 
 
 
 
 
 
 
 
 
 
 
827
 
828#ifdef CONFIG_CEPH_FS_POSIX_ACL
829
830struct posix_acl *ceph_get_acl(struct inode *, int);
831int ceph_set_acl(struct inode *inode, struct posix_acl *acl, int type);
 
832int ceph_pre_init_acls(struct inode *dir, umode_t *mode,
833		       struct ceph_acls_info *info);
834void ceph_init_inode_acls(struct inode *inode, struct ceph_acls_info *info);
835void ceph_release_acls_info(struct ceph_acls_info *info);
836
837static inline void ceph_forget_all_cached_acls(struct inode *inode)
838{
839       forget_all_cached_acls(inode);
840}
841
842#else
843
844#define ceph_get_acl NULL
845#define ceph_set_acl NULL
846
847static inline int ceph_pre_init_acls(struct inode *dir, umode_t *mode,
848				     struct ceph_acls_info *info)
849{
850	return 0;
851}
852static inline void ceph_init_inode_acls(struct inode *inode,
853					struct ceph_acls_info *info)
854{
855}
856static inline void ceph_release_acls_info(struct ceph_acls_info *info)
857{
858}
859static inline int ceph_acl_chmod(struct dentry *dentry, struct inode *inode)
860{
861	return 0;
862}
863
864static inline void ceph_forget_all_cached_acls(struct inode *inode)
865{
866}
867
868#endif
869
870/* caps.c */
871extern const char *ceph_cap_string(int c);
872extern void ceph_handle_caps(struct ceph_mds_session *session,
873			     struct ceph_msg *msg);
874extern struct ceph_cap *ceph_get_cap(struct ceph_mds_client *mdsc,
875				     struct ceph_cap_reservation *ctx);
876extern void ceph_add_cap(struct inode *inode,
877			 struct ceph_mds_session *session, u64 cap_id,
878			 int fmode, unsigned issued, unsigned wanted,
879			 unsigned cap, unsigned seq, u64 realmino, int flags,
880			 struct ceph_cap **new_cap);
881extern void __ceph_remove_cap(struct ceph_cap *cap, bool queue_release);
 
 
 
882extern void ceph_put_cap(struct ceph_mds_client *mdsc,
883			 struct ceph_cap *cap);
884extern int ceph_is_any_caps(struct inode *inode);
885
886extern void ceph_queue_caps_release(struct inode *inode);
887extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc);
888extern int ceph_fsync(struct file *file, loff_t start, loff_t end,
889		      int datasync);
890extern void ceph_early_kick_flushing_caps(struct ceph_mds_client *mdsc,
891					  struct ceph_mds_session *session);
892extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc,
893				    struct ceph_mds_session *session);
 
 
 
 
894extern struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci,
895					     int mds);
896extern int ceph_get_cap_mds(struct inode *inode);
 
897extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps);
898extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had);
 
899extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
900				       struct ceph_snap_context *snapc);
901extern void __ceph_flush_snaps(struct ceph_inode_info *ci,
902			       struct ceph_mds_session **psession,
903			       int again);
904extern void ceph_check_caps(struct ceph_inode_info *ci, int flags,
905			    struct ceph_mds_session *session);
906extern void ceph_check_delayed_caps(struct ceph_mds_client *mdsc);
 
 
 
 
 
907extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc);
908
 
909extern int ceph_encode_inode_release(void **p, struct inode *inode,
910				     int mds, int drop, int unless, int force);
911extern int ceph_encode_dentry_release(void **p, struct dentry *dn,
 
912				      int mds, int drop, int unless);
913
914extern int ceph_get_caps(struct ceph_inode_info *ci, int need, int want,
915			 loff_t endoff, int *got, struct page **pinned_page);
 
 
 
 
916
917/* for counting open files by mode */
918static inline void __ceph_get_fmode(struct ceph_inode_info *ci, int mode)
919{
920	ci->i_nr_by_mode[mode]++;
921}
922extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode);
923
924/* addr.c */
925extern const struct address_space_operations ceph_aops;
 
926extern int ceph_mmap(struct file *file, struct vm_area_struct *vma);
927extern int ceph_uninline_data(struct file *filp, struct page *locked_page);
928extern int ceph_pool_perm_check(struct ceph_inode_info *ci, int need);
929extern void ceph_pool_perm_destroy(struct ceph_mds_client* mdsc);
 
 
 
 
 
 
 
 
 
930
931/* file.c */
932extern const struct file_operations ceph_file_fops;
933
 
934extern int ceph_open(struct inode *inode, struct file *file);
935extern int ceph_atomic_open(struct inode *dir, struct dentry *dentry,
936			    struct file *file, unsigned flags, umode_t mode,
937			    int *opened);
 
 
938extern int ceph_release(struct inode *inode, struct file *filp);
939extern void ceph_fill_inline_data(struct inode *inode, struct page *locked_page,
940				  char *data, size_t len);
 
941/* dir.c */
942extern const struct file_operations ceph_dir_fops;
943extern const struct file_operations ceph_snapdir_fops;
944extern const struct inode_operations ceph_dir_iops;
945extern const struct inode_operations ceph_snapdir_iops;
946extern const struct dentry_operations ceph_dentry_ops, ceph_snap_dentry_ops,
947	ceph_snapdir_dentry_ops;
948
 
949extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry);
950extern int ceph_handle_snapdir(struct ceph_mds_request *req,
951			       struct dentry *dentry, int err);
952extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req,
953					 struct dentry *dentry, int err);
954
955extern void ceph_dentry_lru_add(struct dentry *dn);
956extern void ceph_dentry_lru_touch(struct dentry *dn);
957extern void ceph_dentry_lru_del(struct dentry *dn);
958extern void ceph_invalidate_dentry_lease(struct dentry *dentry);
 
959extern unsigned ceph_dentry_hash(struct inode *dir, struct dentry *dn);
960extern void ceph_readdir_cache_release(struct ceph_readdir_cache_control *ctl);
961
962/*
963 * our d_ops vary depending on whether the inode is live,
964 * snapshotted (read-only), or a virtual ".snap" directory.
965 */
966int ceph_init_dentry(struct dentry *dentry);
967
968
969/* ioctl.c */
970extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
971
972/* export.c */
973extern const struct export_operations ceph_export_ops;
 
974
975/* locks.c */
976extern __init void ceph_flock_init(void);
977extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl);
978extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl);
979extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num);
980extern int ceph_encode_locks_to_buffer(struct inode *inode,
981				       struct ceph_filelock *flocks,
982				       int num_fcntl_locks,
983				       int num_flock_locks);
984extern int ceph_locks_to_pagelist(struct ceph_filelock *flocks,
985				  struct ceph_pagelist *pagelist,
986				  int num_fcntl_locks, int num_flock_locks);
987extern int lock_to_ceph_filelock(struct file_lock *fl, struct ceph_filelock *c);
988
989/* debugfs.c */
990extern int ceph_fs_debugfs_init(struct ceph_fs_client *client);
991extern void ceph_fs_debugfs_cleanup(struct ceph_fs_client *client);
992
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
993#endif /* _FS_CEPH_SUPER_H */
v6.13.7
   1/* SPDX-License-Identifier: GPL-2.0 */
   2#ifndef _FS_CEPH_SUPER_H
   3#define _FS_CEPH_SUPER_H
   4
   5#include <linux/ceph/ceph_debug.h>
   6#include <linux/ceph/osd_client.h>
   7
   8#include <linux/unaligned.h>
   9#include <linux/backing-dev.h>
  10#include <linux/completion.h>
  11#include <linux/exportfs.h>
  12#include <linux/fs.h>
  13#include <linux/mempool.h>
  14#include <linux/pagemap.h>
  15#include <linux/wait.h>
  16#include <linux/writeback.h>
  17#include <linux/slab.h>
  18#include <linux/posix_acl.h>
  19#include <linux/refcount.h>
  20#include <linux/security.h>
  21#include <linux/netfs.h>
 
  22#include <linux/fscache.h>
  23#include <linux/hashtable.h>
  24
  25#include <linux/ceph/libceph.h>
  26#include "crypto.h"
  27
  28/* large granularity for statfs utilization stats to facilitate
  29 * large volume sizes on 32-bit machines. */
  30#define CEPH_BLOCK_SHIFT   22  /* 4 MB */
  31#define CEPH_BLOCK         (1 << CEPH_BLOCK_SHIFT)
  32#define CEPH_4K_BLOCK_SHIFT 12  /* 4 KB */
  33
  34#define CEPH_MOUNT_OPT_CLEANRECOVER    (1<<1) /* auto reonnect (clean mode) after blocklisted */
  35#define CEPH_MOUNT_OPT_DIRSTAT         (1<<4) /* `cat dirname` for stats */
  36#define CEPH_MOUNT_OPT_RBYTES          (1<<5) /* dir st_bytes = rbytes */
  37#define CEPH_MOUNT_OPT_NOASYNCREADDIR  (1<<7) /* no dcache readdir */
  38#define CEPH_MOUNT_OPT_INO32           (1<<8) /* 32 bit inos */
  39#define CEPH_MOUNT_OPT_DCACHE          (1<<9) /* use dcache for readdir etc */
  40#define CEPH_MOUNT_OPT_FSCACHE         (1<<10) /* use fscache */
  41#define CEPH_MOUNT_OPT_NOPOOLPERM      (1<<11) /* no pool permission check */
  42#define CEPH_MOUNT_OPT_MOUNTWAIT       (1<<12) /* mount waits if no mds is up */
  43#define CEPH_MOUNT_OPT_NOQUOTADF       (1<<13) /* no root dir quota in statfs */
  44#define CEPH_MOUNT_OPT_NOCOPYFROM      (1<<14) /* don't use RADOS 'copy-from' op */
  45#define CEPH_MOUNT_OPT_ASYNC_DIROPS    (1<<15) /* allow async directory ops */
  46#define CEPH_MOUNT_OPT_NOPAGECACHE     (1<<16) /* bypass pagecache altogether */
  47#define CEPH_MOUNT_OPT_SPARSEREAD      (1<<17) /* always do sparse reads */
  48
  49#define CEPH_MOUNT_OPT_DEFAULT			\
  50	(CEPH_MOUNT_OPT_DCACHE |		\
  51	 CEPH_MOUNT_OPT_NOCOPYFROM |		\
  52	 CEPH_MOUNT_OPT_ASYNC_DIROPS)
  53
  54#define ceph_set_mount_opt(fsc, opt) \
  55	(fsc)->mount_options->flags |= CEPH_MOUNT_OPT_##opt
  56#define ceph_clear_mount_opt(fsc, opt) \
  57	(fsc)->mount_options->flags &= ~CEPH_MOUNT_OPT_##opt
  58#define ceph_test_mount_opt(fsc, opt) \
  59	(!!((fsc)->mount_options->flags & CEPH_MOUNT_OPT_##opt))
  60
  61/* max size of osd read request, limited by libceph */
  62#define CEPH_MAX_READ_SIZE              CEPH_MSG_MAX_DATA_LEN
  63/* osd has a configurable limitation of max write size.
  64 * CEPH_MSG_MAX_DATA_LEN should be small enough. */
  65#define CEPH_MAX_WRITE_SIZE		CEPH_MSG_MAX_DATA_LEN
  66#define CEPH_RASIZE_DEFAULT             (8192*1024)    /* max readahead */
  67#define CEPH_MAX_READDIR_DEFAULT        1024
  68#define CEPH_MAX_READDIR_BYTES_DEFAULT  (512*1024)
  69#define CEPH_SNAPDIRNAME_DEFAULT        ".snap"
  70
  71/*
  72 * Delay telling the MDS we no longer want caps, in case we reopen
  73 * the file.  Delay a minimum amount of time, even if we send a cap
  74 * message for some other reason.  Otherwise, take the oppotunity to
  75 * update the mds to avoid sending another message later.
  76 */
  77#define CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT      5  /* cap release delay */
  78#define CEPH_CAPS_WANTED_DELAY_MAX_DEFAULT     60  /* cap release delay */
  79
  80struct ceph_mount_options {
  81	unsigned int flags;
 
  82
  83	unsigned int wsize;            /* max write size */
  84	unsigned int rsize;            /* max read size */
  85	unsigned int rasize;           /* max readahead */
  86	unsigned int congestion_kb;    /* max writeback in flight */
  87	unsigned int caps_wanted_delay_min, caps_wanted_delay_max;
  88	int caps_max;
  89	unsigned int max_readdir;       /* max readdir result (entries) */
  90	unsigned int max_readdir_bytes; /* max readdir result (bytes) */
  91
  92	bool new_dev_syntax;
  93
  94	/*
  95	 * everything above this point can be memcmp'd; everything below
  96	 * is handled in compare_mount_options()
  97	 */
  98
  99	char *snapdir_name;   /* default ".snap" */
 100	char *mds_namespace;  /* default NULL */
 101	char *server_path;    /* default NULL (means "/") */
 102	char *fscache_uniq;   /* default NULL */
 103	char *mon_addr;
 104	struct fscrypt_dummy_policy dummy_enc_policy;
 105};
 106
 107/* mount state */
 108enum {
 109	CEPH_MOUNT_MOUNTING,
 110	CEPH_MOUNT_MOUNTED,
 111	CEPH_MOUNT_UNMOUNTING,
 112	CEPH_MOUNT_UNMOUNTED,
 113	CEPH_MOUNT_SHUTDOWN,
 114	CEPH_MOUNT_RECOVER,
 115	CEPH_MOUNT_FENCE_IO,
 116};
 117
 118#define CEPH_ASYNC_CREATE_CONFLICT_BITS 8
 119
 120struct ceph_fs_client {
 121	struct super_block *sb;
 122
 123	struct list_head metric_wakeup;
 124
 125	struct ceph_mount_options *mount_options;
 126	struct ceph_client *client;
 127
 128	int mount_state;
 129
 130	bool blocklisted;
 131
 132	bool have_copy_from2;
 133
 134	u32 filp_gen;
 135	loff_t max_file_size;
 136
 137	struct ceph_mds_client *mdsc;
 138
 
 
 
 
 
 139	atomic_long_t writeback_count;
 140	bool write_congested;
 141
 142	struct workqueue_struct *inode_wq;
 143	struct workqueue_struct *cap_wq;
 144
 145	DECLARE_HASHTABLE(async_unlink_conflict, CEPH_ASYNC_CREATE_CONFLICT_BITS);
 146	spinlock_t async_unlink_conflict_lock;
 147
 148#ifdef CONFIG_DEBUG_FS
 149	struct dentry *debugfs_dentry_lru, *debugfs_caps;
 150	struct dentry *debugfs_congestion_kb;
 151	struct dentry *debugfs_bdi;
 152	struct dentry *debugfs_mdsc, *debugfs_mdsmap;
 153	struct dentry *debugfs_status;
 154	struct dentry *debugfs_mds_sessions;
 155	struct dentry *debugfs_metrics_dir;
 156#endif
 157
 158#ifdef CONFIG_CEPH_FSCACHE
 159	struct fscache_volume *fscache;
 160#endif
 161#ifdef CONFIG_FS_ENCRYPTION
 162	struct fscrypt_dummy_policy fsc_dummy_enc_policy;
 163#endif
 164};
 165
 
 166/*
 167 * File i/o capability.  This tracks shared state with the metadata
 168 * server that allows us to cache or writeback attributes or to read
 169 * and write data.  For any given inode, we should have one or more
 170 * capabilities, one issued by each metadata server, and our
 171 * cumulative access is the OR of all issued capabilities.
 172 *
 173 * Each cap is referenced by the inode's i_caps rbtree and by per-mds
 174 * session capability lists.
 175 */
 176struct ceph_cap {
 177	struct ceph_inode_info *ci;
 178	struct rb_node ci_node;          /* per-ci cap tree */
 179	struct ceph_mds_session *session;
 180	struct list_head session_caps;   /* per-session caplist */
 181	u64 cap_id;       /* unique cap id (mds provided) */
 182	union {
 183		/* in-use caps */
 184		struct {
 185			int issued;       /* latest, from the mds */
 186			int implemented;  /* implemented superset of
 187					     issued (for revocation) */
 188			int mds;	  /* mds index for this cap */
 189			int mds_wanted;   /* caps wanted from this mds */
 190		};
 191		/* caps to release */
 192		struct {
 193			u64 cap_ino;
 194			int queue_release;
 195		};
 196	};
 197	u32 seq, issue_seq, mseq;
 198	u32 cap_gen;      /* active/stale cycle */
 199	unsigned long last_used;
 200	struct list_head caps_item;
 201};
 202
 203#define CHECK_CAPS_AUTHONLY     1  /* only check auth cap */
 204#define CHECK_CAPS_FLUSH        2  /* flush any dirty caps */
 205#define CHECK_CAPS_NOINVAL      4  /* don't invalidate pagecache */
 206#define CHECK_CAPS_FLUSH_FORCE  8  /* force flush any caps */
 207
 208struct ceph_cap_flush {
 209	u64 tid;
 210	int caps;
 211	bool wake; /* wake up flush waiters when finish ? */
 212	bool is_capsnap; /* true means capsnap */
 213	struct list_head g_list; // global
 214	struct list_head i_list; // per inode
 215};
 216
 217/*
 218 * Snapped cap state that is pending flush to mds.  When a snapshot occurs,
 219 * we first complete any in-process sync writes and writeback any dirty
 220 * data before flushing the snapped state (tracked here) back to the MDS.
 221 */
 222struct ceph_cap_snap {
 223	refcount_t nref;
 224	struct list_head ci_item;
 225
 226	struct ceph_cap_flush cap_flush;
 227
 228	u64 follows;
 229	int issued, dirty;
 230	struct ceph_snap_context *context;
 231
 232	umode_t mode;
 233	kuid_t uid;
 234	kgid_t gid;
 235
 236	struct ceph_buffer *xattr_blob;
 237	u64 xattr_version;
 238
 239	u64 size;
 240	u64 change_attr;
 241	struct timespec64 mtime, atime, ctime, btime;
 242	u64 time_warp_seq;
 243	u64 truncate_size;
 244	u32 truncate_seq;
 245	int writing;   /* a sync write is still in progress */
 246	int dirty_pages;     /* dirty pages awaiting writeback */
 247	bool inline_data;
 248	bool need_flush;
 249};
 250
 251static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap)
 252{
 253	if (refcount_dec_and_test(&capsnap->nref)) {
 254		if (capsnap->xattr_blob)
 255			ceph_buffer_put(capsnap->xattr_blob);
 256		kmem_cache_free(ceph_cap_snap_cachep, capsnap);
 257	}
 258}
 259
 
 
 
 
 
 
 
 
 
 
 260/*
 261 * The frag tree describes how a directory is fragmented, potentially across
 262 * multiple metadata servers.  It is also used to indicate points where
 263 * metadata authority is delegated, and whether/where metadata is replicated.
 264 *
 265 * A _leaf_ frag will be present in the i_fragtree IFF there is
 266 * delegation info.  That is, if mds >= 0 || ndist > 0.
 267 */
 268#define CEPH_MAX_DIRFRAG_REP 4
 269
 270struct ceph_inode_frag {
 271	struct rb_node node;
 272
 273	/* fragtree state */
 274	u32 frag;
 275	int split_by;         /* i.e. 2^(split_by) children */
 276
 277	/* delegation and replication info */
 278	int mds;              /* -1 if same authority as parent */
 279	int ndist;            /* >0 if replicated */
 280	int dist[CEPH_MAX_DIRFRAG_REP];
 281};
 282
 283/*
 284 * We cache inode xattrs as an encoded blob until they are first used,
 285 * at which point we parse them into an rbtree.
 286 */
 287struct ceph_inode_xattr {
 288	struct rb_node node;
 289
 290	const char *name;
 291	int name_len;
 292	const char *val;
 293	int val_len;
 294	int dirty;
 295
 296	int should_free_name;
 297	int should_free_val;
 298};
 299
 300/*
 301 * Ceph dentry state
 302 */
 303struct ceph_dentry_info {
 304	struct dentry *dentry;
 305	struct ceph_mds_session *lease_session;
 306	struct list_head lease_list;
 307	struct hlist_node hnode;
 308	unsigned long flags;
 309	int lease_shared_gen;
 310	u32 lease_gen;
 311	u32 lease_seq;
 312	unsigned long lease_renew_after, lease_renew_from;
 313	unsigned long time;
 
 
 314	u64 offset;
 315};
 316
 317#define CEPH_DENTRY_REFERENCED		(1 << 0)
 318#define CEPH_DENTRY_LEASE_LIST		(1 << 1)
 319#define CEPH_DENTRY_SHRINK_LIST		(1 << 2)
 320#define CEPH_DENTRY_PRIMARY_LINK	(1 << 3)
 321#define CEPH_DENTRY_ASYNC_UNLINK_BIT	(4)
 322#define CEPH_DENTRY_ASYNC_UNLINK	(1 << CEPH_DENTRY_ASYNC_UNLINK_BIT)
 323#define CEPH_DENTRY_ASYNC_CREATE_BIT	(5)
 324#define CEPH_DENTRY_ASYNC_CREATE	(1 << CEPH_DENTRY_ASYNC_CREATE_BIT)
 325
 326struct ceph_inode_xattrs_info {
 327	/*
 328	 * (still encoded) xattr blob. we avoid the overhead of parsing
 329	 * this until someone actually calls getxattr, etc.
 330	 *
 331	 * blob->vec.iov_len == 4 implies there are no xattrs; blob ==
 332	 * NULL means we don't know.
 333	*/
 334	struct ceph_buffer *blob, *prealloc_blob;
 335
 336	struct rb_root index;
 337	bool dirty;
 338	int count;
 339	int names_size;
 340	int vals_size;
 341	u64 version, index_version;
 342};
 343
 344/*
 345 * Ceph inode.
 346 */
 347struct ceph_inode_info {
 348	struct netfs_inode netfs; /* Netfslib context and vfs inode */
 349	struct ceph_vino i_vino;   /* ceph ino + snap */
 350
 351	spinlock_t i_ceph_lock;
 352
 353	u64 i_version;
 354	u64 i_inline_version;
 355	u32 i_time_warp_seq;
 356
 357	unsigned long i_ceph_flags;
 358	atomic64_t i_release_count;
 359	atomic64_t i_ordered_count;
 360	atomic64_t i_complete_seq[2];
 361
 362	struct ceph_dir_layout i_dir_layout;
 363	struct ceph_file_layout i_layout;
 364	struct ceph_file_layout i_cached_layout;	// for async creates
 365	char *i_symlink;
 366
 367	/* for dirs */
 368	struct timespec64 i_rctime;
 369	u64 i_rbytes, i_rfiles, i_rsubdirs, i_rsnaps;
 370	u64 i_files, i_subdirs;
 371
 372	/* quotas */
 373	u64 i_max_bytes, i_max_files;
 374
 375	s32 i_dir_pin;
 376
 377	struct rb_root i_fragtree;
 378	int i_fragtree_nsplits;
 379	struct mutex i_fragtree_mutex;
 380
 381	struct ceph_inode_xattrs_info i_xattrs;
 382
 383	/* capabilities.  protected _both_ by i_ceph_lock and cap->session's
 384	 * s_mutex. */
 385	struct rb_root i_caps;           /* cap list */
 386	struct ceph_cap *i_auth_cap;     /* authoritative cap, if any */
 387	unsigned i_dirty_caps, i_flushing_caps;     /* mask of dirtied fields */
 388
 389	/*
 390	 * Link to the auth cap's session's s_cap_dirty list. s_cap_dirty
 391	 * is protected by the mdsc->cap_dirty_lock, but each individual item
 392	 * is also protected by the inode's i_ceph_lock. Walking s_cap_dirty
 393	 * requires the mdsc->cap_dirty_lock. List presence for an item can
 394	 * be tested under the i_ceph_lock. Changing anything requires both.
 395	 */
 396	struct list_head i_dirty_item;
 397
 398	/*
 399	 * Link to session's s_cap_flushing list. Protected in a similar
 400	 * fashion to i_dirty_item, but also by the s_mutex for changes. The
 401	 * s_cap_flushing list can be walked while holding either the s_mutex
 402	 * or msdc->cap_dirty_lock. List presence can also be checked while
 403	 * holding the i_ceph_lock for this inode.
 404	 */
 405	struct list_head i_flushing_item;
 406
 407	/* we need to track cap writeback on a per-cap-bit basis, to allow
 408	 * overlapping, pipelined cap flushes to the mds.  we can probably
 409	 * reduce the tid to 8 bits if we're concerned about inode size. */
 410	struct ceph_cap_flush *i_prealloc_cap_flush;
 411	struct list_head i_cap_flush_list;
 412	wait_queue_head_t i_cap_wq;      /* threads waiting on a capability */
 
 413	unsigned long i_hold_caps_max; /* jiffies */
 414	struct list_head i_cap_delay_list;  /* for delayed cap release to mds */
 415	struct ceph_cap_reservation i_cap_migration_resv;
 416	struct list_head i_cap_snaps;   /* snapped state pending flush to mds */
 417	struct ceph_snap_context *i_head_snapc;  /* set if wr_buffer_head > 0 or
 418						    dirty|flushing caps */
 419	unsigned i_snap_caps;           /* cap bits for snapped files */
 420
 421	unsigned long i_last_rd;
 422	unsigned long i_last_wr;
 423	int i_nr_by_mode[CEPH_FILE_MODE_BITS];  /* open file counts */
 424
 425	struct mutex i_truncate_mutex;
 426	u32 i_truncate_seq;        /* last truncate to smaller size */
 427	u64 i_truncate_size;       /*  and the size we last truncated down to */
 428	int i_truncate_pending;    /*  still need to call vmtruncate */
 429	/*
 430	 * For none fscrypt case it equals to i_truncate_size or it will
 431	 * equals to fscrypt_file_size
 432	 */
 433	u64 i_truncate_pagecache_size;
 434
 435	u64 i_max_size;            /* max file size authorized by mds */
 436	u64 i_reported_size; /* (max_)size reported to or requested of mds */
 437	u64 i_wanted_max_size;     /* offset we'd like to write too */
 438	u64 i_requested_max_size;  /* max_size we've requested */
 439
 440	/* held references to caps */
 441	int i_pin_ref;
 442	int i_rd_ref, i_rdcache_ref, i_wr_ref, i_wb_ref, i_fx_ref;
 443	int i_wrbuffer_ref, i_wrbuffer_ref_head;
 444	atomic_t i_filelock_ref;
 445	atomic_t i_shared_gen;       /* increment each time we get FILE_SHARED */
 446	u32 i_rdcache_gen;      /* incremented each time we get FILE_CACHE. */
 447	u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */
 448
 
 449	struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */
 450	struct list_head i_unsafe_iops;   /* uncommitted mds inode ops */
 451	spinlock_t i_unsafe_lock;
 452
 453	union {
 454		struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */
 455		struct ceph_snapid_map *i_snapid_map; /* snapid -> dev_t */
 456	};
 457	struct list_head i_snap_realm_item;
 458	struct list_head i_snap_flush_item;
 459	struct timespec64 i_btime;
 460	struct timespec64 i_snap_btime;
 461
 462	struct work_struct i_work;
 463	unsigned long  i_work_mask;
 
 
 464
 465#ifdef CONFIG_FS_ENCRYPTION
 466	u32 fscrypt_auth_len;
 467	u32 fscrypt_file_len;
 468	u8 *fscrypt_auth;
 469	u8 *fscrypt_file;
 470#endif
 
 471};
 472
 473struct ceph_netfs_request_data {
 474	int caps;
 475
 476	/*
 477	 * Maximum size of a file readahead request.
 478	 * The fadvise could update the bdi's default ra_pages.
 479	 */
 480	unsigned int file_ra_pages;
 481
 482	/* Set it if fadvise disables file readahead entirely */
 483	bool file_ra_disabled;
 484};
 485
 486static inline struct ceph_inode_info *
 487ceph_inode(const struct inode *inode)
 488{
 489	return container_of(inode, struct ceph_inode_info, netfs.inode);
 490}
 491
 492static inline struct ceph_fs_client *
 493ceph_inode_to_fs_client(const struct inode *inode)
 494{
 495	return (struct ceph_fs_client *)inode->i_sb->s_fs_info;
 496}
 497
 498static inline struct ceph_fs_client *
 499ceph_sb_to_fs_client(const struct super_block *sb)
 500{
 501	return (struct ceph_fs_client *)sb->s_fs_info;
 502}
 503
 504static inline struct ceph_mds_client *
 505ceph_sb_to_mdsc(const struct super_block *sb)
 506{
 507	return (struct ceph_mds_client *)ceph_sb_to_fs_client(sb)->mdsc;
 508}
 509
 510static inline struct ceph_client *
 511ceph_inode_to_client(const struct inode *inode)
 512{
 513	return (struct ceph_client *)ceph_inode_to_fs_client(inode)->client;
 514}
 515
 516static inline struct ceph_vino
 517ceph_vino(const struct inode *inode)
 518{
 519	return ceph_inode(inode)->i_vino;
 520}
 521
 522static inline u32 ceph_ino_to_ino32(u64 vino)
 
 
 
 
 
 
 
 
 523{
 524	u32 ino = vino & 0xffffffff;
 525	ino ^= vino >> 32;
 526	if (!ino)
 527		ino = 2;
 528	return ino;
 529}
 530
 531/*
 532 * Inode numbers in cephfs are 64 bits, but inode->i_ino is 32-bits on
 533 * some arches. We generally do not use this value inside the ceph driver, but
 534 * we do want to set it to something, so that generic vfs code has an
 535 * appropriate value for tracepoints and the like.
 536 */
 537static inline ino_t ceph_vino_to_ino_t(struct ceph_vino vino)
 538{
 539	if (sizeof(ino_t) == sizeof(u32))
 540		return ceph_ino_to_ino32(vino.ino);
 
 541	return (ino_t)vino.ino;
 
 542}
 543
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 544/* for printf-style formatting */
 545#define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap
 546
 547static inline u64 ceph_ino(struct inode *inode)
 548{
 549	return ceph_inode(inode)->i_vino.ino;
 550}
 551
 552static inline u64 ceph_snap(struct inode *inode)
 553{
 554	return ceph_inode(inode)->i_vino.snap;
 555}
 556
 557/**
 558 * ceph_present_ino - format an inode number for presentation to userland
 559 * @sb: superblock where the inode lives
 560 * @ino: inode number to (possibly) convert
 561 *
 562 * If the user mounted with the ino32 option, then the 64-bit value needs
 563 * to be converted to something that can fit inside 32 bits. Note that
 564 * internal kernel code never uses this value, so this is entirely for
 565 * userland consumption.
 566 */
 567static inline u64 ceph_present_ino(struct super_block *sb, u64 ino)
 568{
 569	if (unlikely(ceph_test_mount_opt(ceph_sb_to_fs_client(sb), INO32)))
 570		return ceph_ino_to_ino32(ino);
 571	return ino;
 572}
 573
 574static inline u64 ceph_present_inode(struct inode *inode)
 575{
 576	return ceph_present_ino(inode->i_sb, ceph_ino(inode));
 577}
 578
 579static inline int ceph_ino_compare(struct inode *inode, void *data)
 580{
 581	struct ceph_vino *pvino = (struct ceph_vino *)data;
 582	struct ceph_inode_info *ci = ceph_inode(inode);
 583	return ci->i_vino.ino == pvino->ino &&
 584		ci->i_vino.snap == pvino->snap;
 585}
 586
 587/*
 588 * The MDS reserves a set of inodes for its own usage. These should never
 589 * be accessible by clients, and so the MDS has no reason to ever hand these
 590 * out. The range is CEPH_MDS_INO_MDSDIR_OFFSET..CEPH_INO_SYSTEM_BASE.
 591 *
 592 * These come from src/mds/mdstypes.h in the ceph sources.
 593 */
 594#define CEPH_MAX_MDS			0x100
 595#define CEPH_NUM_STRAY			10
 596#define CEPH_MDS_INO_MDSDIR_OFFSET	(1 * CEPH_MAX_MDS)
 597#define CEPH_MDS_INO_LOG_OFFSET		(2 * CEPH_MAX_MDS)
 598#define CEPH_INO_SYSTEM_BASE		((6*CEPH_MAX_MDS) + (CEPH_MAX_MDS * CEPH_NUM_STRAY))
 599
 600static inline bool ceph_vino_is_reserved(const struct ceph_vino vino)
 601{
 602	if (vino.ino >= CEPH_INO_SYSTEM_BASE ||
 603	    vino.ino < CEPH_MDS_INO_MDSDIR_OFFSET)
 604		return false;
 605
 606	/* Don't warn on mdsdirs */
 607	WARN_RATELIMIT(vino.ino >= CEPH_MDS_INO_LOG_OFFSET,
 608			"Attempt to access reserved inode number 0x%llx",
 609			vino.ino);
 610	return true;
 611}
 612
 613static inline struct inode *ceph_find_inode(struct super_block *sb,
 614					    struct ceph_vino vino)
 615{
 616	if (ceph_vino_is_reserved(vino))
 617		return NULL;
 618
 619	/*
 620	 * NB: The hashval will be run through the fs/inode.c hash function
 621	 * anyway, so there is no need to squash the inode number down to
 622	 * 32-bits first. Just use low-order bits on arches with 32-bit long.
 623	 */
 624	return ilookup5(sb, (unsigned long)vino.ino, ceph_ino_compare, &vino);
 625}
 626
 627
 628/*
 629 * Ceph inode.
 630 */
 631#define CEPH_I_DIR_ORDERED	(1 << 0)  /* dentries in dir are ordered */
 
 632#define CEPH_I_FLUSH		(1 << 2)  /* do not delay flush of dirty metadata */
 633#define CEPH_I_POOL_PERM	(1 << 3)  /* pool rd/wr bits are valid */
 634#define CEPH_I_POOL_RD		(1 << 4)  /* can read from pool */
 635#define CEPH_I_POOL_WR		(1 << 5)  /* can write to pool */
 636#define CEPH_I_SEC_INITED	(1 << 6)  /* security initialized */
 637#define CEPH_I_KICK_FLUSH	(1 << 7)  /* kick flushing caps */
 638#define CEPH_I_FLUSH_SNAPS	(1 << 8)  /* need flush snapss */
 639#define CEPH_I_ERROR_WRITE	(1 << 9) /* have seen write errors */
 640#define CEPH_I_ERROR_FILELOCK	(1 << 10) /* have seen file lock errors */
 641#define CEPH_I_ODIRECT		(1 << 11) /* inode in direct I/O mode */
 642#define CEPH_ASYNC_CREATE_BIT	(12)	  /* async create in flight for this */
 643#define CEPH_I_ASYNC_CREATE	(1 << CEPH_ASYNC_CREATE_BIT)
 644#define CEPH_I_SHUTDOWN		(1 << 13) /* inode is no longer usable */
 645#define CEPH_I_ASYNC_CHECK_CAPS	(1 << 14) /* check caps immediately after async
 646					     creating finishes */
 647
 648/*
 649 * Masks of ceph inode work.
 650 */
 651#define CEPH_I_WORK_WRITEBACK		0
 652#define CEPH_I_WORK_INVALIDATE_PAGES	1
 653#define CEPH_I_WORK_VMTRUNCATE		2
 654#define CEPH_I_WORK_CHECK_CAPS		3
 655#define CEPH_I_WORK_FLUSH_SNAPS		4
 656
 657/*
 658 * We set the ERROR_WRITE bit when we start seeing write errors on an inode
 659 * and then clear it when they start succeeding. Note that we do a lockless
 660 * check first, and only take the lock if it looks like it needs to be changed.
 661 * The write submission code just takes this as a hint, so we're not too
 662 * worried if a few slip through in either direction.
 663 */
 664static inline void ceph_set_error_write(struct ceph_inode_info *ci)
 665{
 666	if (!(READ_ONCE(ci->i_ceph_flags) & CEPH_I_ERROR_WRITE)) {
 667		spin_lock(&ci->i_ceph_lock);
 668		ci->i_ceph_flags |= CEPH_I_ERROR_WRITE;
 669		spin_unlock(&ci->i_ceph_lock);
 670	}
 671}
 672
 673static inline void ceph_clear_error_write(struct ceph_inode_info *ci)
 674{
 675	if (READ_ONCE(ci->i_ceph_flags) & CEPH_I_ERROR_WRITE) {
 676		spin_lock(&ci->i_ceph_lock);
 677		ci->i_ceph_flags &= ~CEPH_I_ERROR_WRITE;
 678		spin_unlock(&ci->i_ceph_lock);
 679	}
 680}
 681
 682static inline void __ceph_dir_set_complete(struct ceph_inode_info *ci,
 683					   long long release_count,
 684					   long long ordered_count)
 685{
 686	/*
 687	 * Makes sure operations that setup readdir cache (update page
 688	 * cache and i_size) are strongly ordered w.r.t. the following
 689	 * atomic64_set() operations.
 690	 */
 691	smp_mb();
 692	atomic64_set(&ci->i_complete_seq[0], release_count);
 693	atomic64_set(&ci->i_complete_seq[1], ordered_count);
 694}
 695
 696static inline void __ceph_dir_clear_complete(struct ceph_inode_info *ci)
 697{
 698	atomic64_inc(&ci->i_release_count);
 699}
 700
 701static inline void __ceph_dir_clear_ordered(struct ceph_inode_info *ci)
 702{
 703	atomic64_inc(&ci->i_ordered_count);
 704}
 705
 706static inline bool __ceph_dir_is_complete(struct ceph_inode_info *ci)
 707{
 708	return atomic64_read(&ci->i_complete_seq[0]) ==
 709		atomic64_read(&ci->i_release_count);
 710}
 711
 712static inline bool __ceph_dir_is_complete_ordered(struct ceph_inode_info *ci)
 713{
 714	return  atomic64_read(&ci->i_complete_seq[0]) ==
 715		atomic64_read(&ci->i_release_count) &&
 716		atomic64_read(&ci->i_complete_seq[1]) ==
 717		atomic64_read(&ci->i_ordered_count);
 718}
 719
 720static inline void ceph_dir_clear_complete(struct inode *inode)
 721{
 722	__ceph_dir_clear_complete(ceph_inode(inode));
 723}
 724
 725static inline void ceph_dir_clear_ordered(struct inode *inode)
 726{
 727	__ceph_dir_clear_ordered(ceph_inode(inode));
 728}
 729
 730static inline bool ceph_dir_is_complete_ordered(struct inode *inode)
 731{
 732	bool ret = __ceph_dir_is_complete_ordered(ceph_inode(inode));
 733	smp_rmb();
 734	return ret;
 735}
 736
 737/* find a specific frag @f */
 738extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci,
 739						u32 f);
 740
 741/*
 742 * choose fragment for value @v.  copy frag content to pfrag, if leaf
 743 * exists
 744 */
 745extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v,
 746			    struct ceph_inode_frag *pfrag,
 747			    int *found);
 748
 749static inline struct ceph_dentry_info *ceph_dentry(const struct dentry *dentry)
 750{
 751	return (struct ceph_dentry_info *)dentry->d_fsdata;
 752}
 753
 
 
 
 
 
 754/*
 755 * caps helpers
 756 */
 757static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci)
 758{
 759	return !RB_EMPTY_ROOT(&ci->i_caps);
 760}
 761
 762extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented);
 763extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t);
 764extern int __ceph_caps_issued_mask_metric(struct ceph_inode_info *ci, int mask,
 765					  int t);
 766extern int __ceph_caps_issued_other(struct ceph_inode_info *ci,
 767				    struct ceph_cap *cap);
 768
 769static inline int ceph_caps_issued(struct ceph_inode_info *ci)
 770{
 771	int issued;
 772	spin_lock(&ci->i_ceph_lock);
 773	issued = __ceph_caps_issued(ci, NULL);
 774	spin_unlock(&ci->i_ceph_lock);
 775	return issued;
 776}
 777
 778static inline int ceph_caps_issued_mask_metric(struct ceph_inode_info *ci,
 779					       int mask, int touch)
 780{
 781	int r;
 782	spin_lock(&ci->i_ceph_lock);
 783	r = __ceph_caps_issued_mask_metric(ci, mask, touch);
 784	spin_unlock(&ci->i_ceph_lock);
 785	return r;
 786}
 787
 788static inline int __ceph_caps_dirty(struct ceph_inode_info *ci)
 789{
 790	return ci->i_dirty_caps | ci->i_flushing_caps;
 791}
 792extern struct ceph_cap_flush *ceph_alloc_cap_flush(void);
 793extern void ceph_free_cap_flush(struct ceph_cap_flush *cf);
 794extern int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask,
 795				  struct ceph_cap_flush **pcf);
 796
 797extern int __ceph_caps_revoking_other(struct ceph_inode_info *ci,
 798				      struct ceph_cap *ocap, int mask);
 
 799extern int __ceph_caps_used(struct ceph_inode_info *ci);
 800
 801static inline bool __ceph_is_file_opened(struct ceph_inode_info *ci)
 
 
 
 
 
 802{
 803	return ci->i_nr_by_mode[0];
 
 
 
 804}
 805extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci);
 806extern int __ceph_caps_wanted(struct ceph_inode_info *ci);
 807
 808/* what the mds thinks we want */
 809extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci, bool check);
 810
 811extern void ceph_caps_init(struct ceph_mds_client *mdsc);
 812extern void ceph_caps_finalize(struct ceph_mds_client *mdsc);
 813extern void ceph_adjust_caps_max_min(struct ceph_mds_client *mdsc,
 814				     struct ceph_mount_options *fsopt);
 815extern int ceph_reserve_caps(struct ceph_mds_client *mdsc,
 816			     struct ceph_cap_reservation *ctx, int need);
 817extern void ceph_unreserve_caps(struct ceph_mds_client *mdsc,
 818			       struct ceph_cap_reservation *ctx);
 819extern void ceph_reservation_status(struct ceph_fs_client *client,
 820				    int *total, int *avail, int *used,
 821				    int *reserved, int *min);
 822extern void change_auth_cap_ses(struct ceph_inode_info *ci,
 823				struct ceph_mds_session *session);
 824
 825
 826
 827/*
 828 * we keep buffered readdir results attached to file->private_data
 829 */
 830#define CEPH_F_SYNC     1
 831#define CEPH_F_ATEND    2
 832
 833struct ceph_file_info {
 834	short fmode;     /* initialized on open */
 835	short flags;     /* CEPH_F_* */
 836
 837	spinlock_t rw_contexts_lock;
 838	struct list_head rw_contexts;
 839
 840	u32 filp_gen;
 841};
 842
 843struct ceph_dir_file_info {
 844	struct ceph_file_info file_info;
 845
 846	/* readdir: position within the dir */
 847	u32 frag;
 848	struct ceph_mds_request *last_readdir;
 849
 850	/* readdir: position within a frag */
 
 851	unsigned next_offset;  /* offset of next chunk (last_name's + 1) */
 852	char *last_name;       /* last entry in previous chunk */
 853	long long dir_release_count;
 854	long long dir_ordered_count;
 855	int readdir_cache_idx;
 856
 857	/* used for -o dirstat read() on directory thing */
 858	char *dir_info;
 859	int dir_info_len;
 860};
 861
 862struct ceph_rw_context {
 863	struct list_head list;
 864	struct task_struct *thread;
 865	int caps;
 866};
 867
 868#define CEPH_DEFINE_RW_CONTEXT(_name, _caps)	\
 869	struct ceph_rw_context _name = {	\
 870		.thread = current,		\
 871		.caps = _caps,			\
 872	}
 873
 874static inline void ceph_add_rw_context(struct ceph_file_info *cf,
 875				       struct ceph_rw_context *ctx)
 876{
 877	spin_lock(&cf->rw_contexts_lock);
 878	list_add(&ctx->list, &cf->rw_contexts);
 879	spin_unlock(&cf->rw_contexts_lock);
 880}
 881
 882static inline void ceph_del_rw_context(struct ceph_file_info *cf,
 883				       struct ceph_rw_context *ctx)
 884{
 885	spin_lock(&cf->rw_contexts_lock);
 886	list_del(&ctx->list);
 887	spin_unlock(&cf->rw_contexts_lock);
 888}
 889
 890static inline struct ceph_rw_context*
 891ceph_find_rw_context(struct ceph_file_info *cf)
 892{
 893	struct ceph_rw_context *ctx, *found = NULL;
 894	spin_lock(&cf->rw_contexts_lock);
 895	list_for_each_entry(ctx, &cf->rw_contexts, list) {
 896		if (ctx->thread == current) {
 897			found = ctx;
 898			break;
 899		}
 900	}
 901	spin_unlock(&cf->rw_contexts_lock);
 902	return found;
 903}
 904
 905struct ceph_readdir_cache_control {
 906	struct page  *page;
 907	struct dentry **dentries;
 908	int index;
 909};
 910
 911/*
 912 * A "snap realm" describes a subset of the file hierarchy sharing
 913 * the same set of snapshots that apply to it.  The realms themselves
 914 * are organized into a hierarchy, such that children inherit (some of)
 915 * the snapshots of their parents.
 916 *
 917 * All inodes within the realm that have capabilities are linked into a
 918 * per-realm list.
 919 */
 920struct ceph_snap_realm {
 921	u64 ino;
 922	struct inode *inode;
 923	atomic_t nref;
 924	struct rb_node node;
 925
 926	u64 created, seq;
 927	u64 parent_ino;
 928	u64 parent_since;   /* snapid when our current parent became so */
 929
 930	u64 *prior_parent_snaps;      /* snaps inherited from any parents we */
 931	u32 num_prior_parent_snaps;   /*  had prior to parent_since */
 932	u64 *snaps;                   /* snaps specific to this realm */
 933	u32 num_snaps;
 934
 935	struct ceph_snap_realm *parent;
 936	struct list_head children;       /* list of child realms */
 937	struct list_head child_item;
 938
 939	struct list_head empty_item;     /* if i have ref==0 */
 940
 941	struct list_head dirty_item;     /* if realm needs new context */
 942
 943	struct list_head rebuild_item;   /* rebuild snap realms _downward_ in hierarchy */
 944
 945	/* the current set of snaps for this realm */
 946	struct ceph_snap_context *cached_context;
 947
 948	struct list_head inodes_with_caps;
 949	spinlock_t inodes_with_caps_lock;
 950};
 951
 952static inline int default_congestion_kb(void)
 953{
 954	int congestion_kb;
 955
 956	/*
 957	 * Copied from NFS
 958	 *
 959	 * congestion size, scale with available memory.
 960	 *
 961	 *  64MB:    8192k
 962	 * 128MB:   11585k
 963	 * 256MB:   16384k
 964	 * 512MB:   23170k
 965	 *   1GB:   32768k
 966	 *   2GB:   46340k
 967	 *   4GB:   65536k
 968	 *   8GB:   92681k
 969	 *  16GB:  131072k
 970	 *
 971	 * This allows larger machines to have larger/more transfers.
 972	 * Limit the default to 256M
 973	 */
 974	congestion_kb = (16*int_sqrt(totalram_pages())) << (PAGE_SHIFT-10);
 975	if (congestion_kb > 256*1024)
 976		congestion_kb = 256*1024;
 977
 978	return congestion_kb;
 979}
 980
 981
 982/* super.c */
 983extern int ceph_force_reconnect(struct super_block *sb);
 984/* snap.c */
 985struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc,
 986					       u64 ino);
 987extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc,
 988				struct ceph_snap_realm *realm);
 989extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc,
 990				struct ceph_snap_realm *realm);
 991extern int ceph_update_snap_trace(struct ceph_mds_client *m,
 992				  void *p, void *e, bool deletion,
 993				  struct ceph_snap_realm **realm_ret);
 994void ceph_change_snap_realm(struct inode *inode, struct ceph_snap_realm *realm);
 995extern void ceph_handle_snap(struct ceph_mds_client *mdsc,
 996			     struct ceph_mds_session *session,
 997			     struct ceph_msg *msg);
 
 998extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci,
 999				  struct ceph_cap_snap *capsnap);
1000extern void ceph_cleanup_global_and_empty_realms(struct ceph_mds_client *mdsc);
1001
1002extern struct ceph_snapid_map *ceph_get_snapid_map(struct ceph_mds_client *mdsc,
1003						   u64 snap);
1004extern void ceph_put_snapid_map(struct ceph_mds_client* mdsc,
1005				struct ceph_snapid_map *sm);
1006extern void ceph_trim_snapid_map(struct ceph_mds_client *mdsc);
1007extern void ceph_cleanup_snapid_map(struct ceph_mds_client *mdsc);
1008void ceph_umount_begin(struct super_block *sb);
1009
1010
1011/*
1012 * a cap_snap is "pending" if it is still awaiting an in-progress
1013 * sync write (that may/may not still update size, mtime, etc.).
1014 */
1015static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci)
1016{
1017	return !list_empty(&ci->i_cap_snaps) &&
1018	       list_last_entry(&ci->i_cap_snaps, struct ceph_cap_snap,
1019			       ci_item)->writing;
1020}
1021
1022/* inode.c */
1023struct ceph_mds_reply_info_in;
1024struct ceph_mds_reply_dirfrag;
1025struct ceph_acl_sec_ctx;
1026
1027extern const struct inode_operations ceph_file_iops;
1028
1029extern struct inode *ceph_alloc_inode(struct super_block *sb);
1030extern void ceph_evict_inode(struct inode *inode);
1031extern void ceph_free_inode(struct inode *inode);
1032
1033struct inode *ceph_new_inode(struct inode *dir, struct dentry *dentry,
1034			     umode_t *mode, struct ceph_acl_sec_ctx *as_ctx);
1035void ceph_as_ctx_to_req(struct ceph_mds_request *req,
1036			struct ceph_acl_sec_ctx *as_ctx);
1037
1038extern struct inode *ceph_get_inode(struct super_block *sb,
1039				    struct ceph_vino vino,
1040				    struct inode *newino);
1041extern struct inode *ceph_get_snapdir(struct inode *parent);
1042extern int ceph_fill_file_size(struct inode *inode, int issued,
1043			       u32 truncate_seq, u64 truncate_size, u64 size);
1044extern void ceph_fill_file_time(struct inode *inode, int issued,
1045				u64 time_warp_seq, struct timespec64 *ctime,
1046				struct timespec64 *mtime,
1047				struct timespec64 *atime);
1048extern int ceph_fill_inode(struct inode *inode, struct page *locked_page,
1049		    struct ceph_mds_reply_info_in *iinfo,
1050		    struct ceph_mds_reply_dirfrag *dirinfo,
1051		    struct ceph_mds_session *session, int cap_fmode,
1052		    struct ceph_cap_reservation *caps_reservation);
1053extern int ceph_fill_trace(struct super_block *sb,
1054			   struct ceph_mds_request *req);
 
1055extern int ceph_readdir_prepopulate(struct ceph_mds_request *req,
1056				    struct ceph_mds_session *session);
1057
1058extern bool ceph_inode_set_size(struct inode *inode, loff_t size);
 
 
1059extern void __ceph_do_pending_vmtruncate(struct inode *inode);
 
1060
1061void ceph_queue_inode_work(struct inode *inode, int work_bit);
1062
1063static inline void ceph_queue_vmtruncate(struct inode *inode)
1064{
1065	ceph_queue_inode_work(inode, CEPH_I_WORK_VMTRUNCATE);
1066}
1067
1068static inline void ceph_queue_invalidate(struct inode *inode)
1069{
1070	ceph_queue_inode_work(inode, CEPH_I_WORK_INVALIDATE_PAGES);
1071}
1072
1073static inline void ceph_queue_writeback(struct inode *inode)
1074{
1075	ceph_queue_inode_work(inode, CEPH_I_WORK_WRITEBACK);
1076}
1077
1078static inline void ceph_queue_check_caps(struct inode *inode)
1079{
1080	ceph_queue_inode_work(inode, CEPH_I_WORK_CHECK_CAPS);
1081}
1082
1083static inline void ceph_queue_flush_snaps(struct inode *inode)
1084{
1085	ceph_queue_inode_work(inode, CEPH_I_WORK_FLUSH_SNAPS);
1086}
1087
1088extern int ceph_try_to_choose_auth_mds(struct inode *inode, int mask);
1089extern int __ceph_do_getattr(struct inode *inode, struct page *locked_page,
1090			     int mask, bool force);
1091static inline int ceph_do_getattr(struct inode *inode, int mask, bool force)
1092{
1093	return __ceph_do_getattr(inode, NULL, mask, force);
1094}
1095extern int ceph_permission(struct mnt_idmap *idmap,
1096			   struct inode *inode, int mask);
1097
1098struct ceph_iattr {
1099	struct ceph_fscrypt_auth	*fscrypt_auth;
1100};
1101
1102extern int __ceph_setattr(struct mnt_idmap *idmap, struct inode *inode,
1103			  struct iattr *attr, struct ceph_iattr *cia);
1104extern int ceph_setattr(struct mnt_idmap *idmap,
1105			struct dentry *dentry, struct iattr *attr);
1106extern int ceph_getattr(struct mnt_idmap *idmap,
1107			const struct path *path, struct kstat *stat,
1108			u32 request_mask, unsigned int flags);
1109void ceph_inode_shutdown(struct inode *inode);
1110
1111static inline bool ceph_inode_is_shutdown(struct inode *inode)
1112{
1113	unsigned long flags = READ_ONCE(ceph_inode(inode)->i_ceph_flags);
1114	struct ceph_fs_client *fsc = ceph_inode_to_fs_client(inode);
1115	int state = READ_ONCE(fsc->mount_state);
1116
1117	return (flags & CEPH_I_SHUTDOWN) || state >= CEPH_MOUNT_SHUTDOWN;
1118}
1119
1120/* xattr.c */
1121int __ceph_setxattr(struct inode *, const char *, const void *, size_t, int);
1122int ceph_do_getvxattr(struct inode *inode, const char *name, void *value, size_t size);
 
1123ssize_t __ceph_getxattr(struct inode *, const char *, void *, size_t);
 
 
1124extern ssize_t ceph_listxattr(struct dentry *, char *, size_t);
1125extern struct ceph_buffer *__ceph_build_xattrs_blob(struct ceph_inode_info *ci);
 
1126extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci);
1127extern const struct xattr_handler * const ceph_xattr_handlers[];
1128
1129struct ceph_acl_sec_ctx {
1130#ifdef CONFIG_CEPH_FS_POSIX_ACL
1131	void *default_acl;
1132	void *acl;
1133#endif
1134#ifdef CONFIG_CEPH_FS_SECURITY_LABEL
1135	void *sec_ctx;
1136	u32 sec_ctxlen;
1137#endif
1138#ifdef CONFIG_FS_ENCRYPTION
1139	struct ceph_fscrypt_auth *fscrypt_auth;
1140#endif
1141	struct ceph_pagelist *pagelist;
1142};
1143
1144#ifdef CONFIG_SECURITY
1145extern bool ceph_security_xattr_deadlock(struct inode *in);
1146extern bool ceph_security_xattr_wanted(struct inode *in);
1147#else
1148static inline bool ceph_security_xattr_deadlock(struct inode *in)
1149{
1150	return false;
1151}
1152static inline bool ceph_security_xattr_wanted(struct inode *in)
1153{
1154	return false;
1155}
1156#endif
1157
1158#ifdef CONFIG_CEPH_FS_SECURITY_LABEL
1159extern int ceph_security_init_secctx(struct dentry *dentry, umode_t mode,
1160				     struct ceph_acl_sec_ctx *ctx);
1161static inline void ceph_security_invalidate_secctx(struct inode *inode)
1162{
1163	security_inode_invalidate_secctx(inode);
1164}
1165#else
1166static inline int ceph_security_init_secctx(struct dentry *dentry, umode_t mode,
1167					    struct ceph_acl_sec_ctx *ctx)
1168{
1169	return 0;
1170}
1171static inline void ceph_security_invalidate_secctx(struct inode *inode)
1172{
1173}
1174#endif
1175
1176void ceph_release_acl_sec_ctx(struct ceph_acl_sec_ctx *as_ctx);
1177
1178/* acl.c */
1179#ifdef CONFIG_CEPH_FS_POSIX_ACL
1180
1181struct posix_acl *ceph_get_acl(struct inode *, int, bool);
1182int ceph_set_acl(struct mnt_idmap *idmap,
1183		 struct dentry *dentry, struct posix_acl *acl, int type);
1184int ceph_pre_init_acls(struct inode *dir, umode_t *mode,
1185		       struct ceph_acl_sec_ctx *as_ctx);
1186void ceph_init_inode_acls(struct inode *inode,
1187			  struct ceph_acl_sec_ctx *as_ctx);
1188
1189static inline void ceph_forget_all_cached_acls(struct inode *inode)
1190{
1191       forget_all_cached_acls(inode);
1192}
1193
1194#else
1195
1196#define ceph_get_acl NULL
1197#define ceph_set_acl NULL
1198
1199static inline int ceph_pre_init_acls(struct inode *dir, umode_t *mode,
1200				     struct ceph_acl_sec_ctx *as_ctx)
1201{
1202	return 0;
1203}
1204static inline void ceph_init_inode_acls(struct inode *inode,
1205					struct ceph_acl_sec_ctx *as_ctx)
1206{
1207}
 
 
 
 
 
 
 
1208
1209static inline void ceph_forget_all_cached_acls(struct inode *inode)
1210{
1211}
1212
1213#endif
1214
1215/* caps.c */
1216extern const char *ceph_cap_string(int c);
1217extern void ceph_handle_caps(struct ceph_mds_session *session,
1218			     struct ceph_msg *msg);
1219extern struct ceph_cap *ceph_get_cap(struct ceph_mds_client *mdsc,
1220				     struct ceph_cap_reservation *ctx);
1221extern void ceph_add_cap(struct inode *inode,
1222			 struct ceph_mds_session *session, u64 cap_id,
1223			 unsigned issued, unsigned wanted,
1224			 unsigned cap, unsigned seq, u64 realmino, int flags,
1225			 struct ceph_cap **new_cap);
1226extern void __ceph_remove_cap(struct ceph_cap *cap, bool queue_release);
1227extern void ceph_remove_cap(struct ceph_mds_client *mdsc, struct ceph_cap *cap,
1228			    bool queue_release);
1229extern void __ceph_remove_caps(struct ceph_inode_info *ci);
1230extern void ceph_put_cap(struct ceph_mds_client *mdsc,
1231			 struct ceph_cap *cap);
1232extern int ceph_is_any_caps(struct inode *inode);
1233
 
1234extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc);
1235extern int ceph_fsync(struct file *file, loff_t start, loff_t end,
1236		      int datasync);
1237extern void ceph_early_kick_flushing_caps(struct ceph_mds_client *mdsc,
1238					  struct ceph_mds_session *session);
1239extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc,
1240				    struct ceph_mds_session *session);
1241void ceph_kick_flushing_inode_caps(struct ceph_mds_session *session,
1242				   struct ceph_inode_info *ci);
1243extern struct ceph_cap *__get_cap_for_mds(struct ceph_inode_info *ci,
1244					  int mds);
1245extern struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci,
1246					     int mds);
1247extern void ceph_take_cap_refs(struct ceph_inode_info *ci, int caps,
1248				bool snap_rwsem_locked);
1249extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps);
1250extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had);
1251extern void ceph_put_cap_refs_async(struct ceph_inode_info *ci, int had);
1252extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
1253				       struct ceph_snap_context *snapc);
1254extern void __ceph_remove_capsnap(struct inode *inode,
1255				  struct ceph_cap_snap *capsnap,
1256				  bool *wake_ci, bool *wake_mdsc);
1257extern void ceph_remove_capsnap(struct inode *inode,
1258				struct ceph_cap_snap *capsnap,
1259				bool *wake_ci, bool *wake_mdsc);
1260extern void ceph_flush_snaps(struct ceph_inode_info *ci,
1261			     struct ceph_mds_session **psession);
1262extern bool __ceph_should_report_size(struct ceph_inode_info *ci);
1263extern void ceph_check_caps(struct ceph_inode_info *ci, int flags);
1264extern unsigned long ceph_check_delayed_caps(struct ceph_mds_client *mdsc);
1265extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc);
1266extern void ceph_flush_cap_releases(struct ceph_mds_client *mdsc);
1267extern int  ceph_drop_caps_for_unlink(struct inode *inode);
1268extern int ceph_encode_inode_release(void **p, struct inode *inode,
1269				     int mds, int drop, int unless, int force);
1270extern int ceph_encode_dentry_release(void **p, struct dentry *dn,
1271				      struct inode *dir,
1272				      int mds, int drop, int unless);
1273
1274extern int __ceph_get_caps(struct inode *inode, struct ceph_file_info *fi,
1275			   int need, int want, loff_t endoff, int *got);
1276extern int ceph_get_caps(struct file *filp, int need, int want,
1277			 loff_t endoff, int *got);
1278extern int ceph_try_get_caps(struct inode *inode,
1279			     int need, int want, bool nonblock, int *got);
1280
1281/* for counting open files by mode */
1282extern void ceph_get_fmode(struct ceph_inode_info *ci, int mode, int count);
1283extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode, int count);
1284extern void __ceph_touch_fmode(struct ceph_inode_info *ci,
1285			       struct ceph_mds_client *mdsc, int fmode);
 
1286
1287/* addr.c */
1288extern const struct address_space_operations ceph_aops;
1289extern const struct netfs_request_ops ceph_netfs_ops;
1290extern int ceph_mmap(struct file *file, struct vm_area_struct *vma);
1291extern int ceph_uninline_data(struct file *file);
1292extern int ceph_pool_perm_check(struct inode *inode, int need);
1293extern void ceph_pool_perm_destroy(struct ceph_mds_client* mdsc);
1294int ceph_purge_inode_cap(struct inode *inode, struct ceph_cap *cap, bool *invalidate);
1295
1296static inline bool ceph_has_inline_data(struct ceph_inode_info *ci)
1297{
1298	if (ci->i_inline_version == CEPH_INLINE_NONE ||
1299	    ci->i_inline_version == 1) /* initial version, no data */
1300		return false;
1301	return true;
1302}
1303
1304/* file.c */
1305extern const struct file_operations ceph_file_fops;
1306
1307extern int ceph_renew_caps(struct inode *inode, int fmode);
1308extern int ceph_open(struct inode *inode, struct file *file);
1309extern int ceph_atomic_open(struct inode *dir, struct dentry *dentry,
1310			    struct file *file, unsigned flags, umode_t mode);
1311extern ssize_t __ceph_sync_read(struct inode *inode, loff_t *ki_pos,
1312				struct iov_iter *to, int *retry_op,
1313				u64 *last_objver);
1314extern int ceph_release(struct inode *inode, struct file *filp);
1315extern void ceph_fill_inline_data(struct inode *inode, struct page *locked_page,
1316				  char *data, size_t len);
1317
1318/* dir.c */
1319extern const struct file_operations ceph_dir_fops;
1320extern const struct file_operations ceph_snapdir_fops;
1321extern const struct inode_operations ceph_dir_iops;
1322extern const struct inode_operations ceph_snapdir_iops;
1323extern const struct dentry_operations ceph_dentry_ops;
 
1324
1325extern loff_t ceph_make_fpos(unsigned high, unsigned off, bool hash_order);
1326extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry);
1327extern struct dentry *ceph_handle_snapdir(struct ceph_mds_request *req,
1328			       struct dentry *dentry);
1329extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req,
1330					 struct dentry *dentry, int err);
1331
1332extern void __ceph_dentry_lease_touch(struct ceph_dentry_info *di);
1333extern void __ceph_dentry_dir_lease_touch(struct ceph_dentry_info *di);
 
1334extern void ceph_invalidate_dentry_lease(struct dentry *dentry);
1335extern int ceph_trim_dentries(struct ceph_mds_client *mdsc);
1336extern unsigned ceph_dentry_hash(struct inode *dir, struct dentry *dn);
1337extern void ceph_readdir_cache_release(struct ceph_readdir_cache_control *ctl);
1338
 
 
 
 
 
 
 
1339/* ioctl.c */
1340extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
1341
1342/* export.c */
1343extern const struct export_operations ceph_export_ops;
1344struct inode *ceph_lookup_inode(struct super_block *sb, u64 ino);
1345
1346/* locks.c */
1347extern __init void ceph_flock_init(void);
1348extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl);
1349extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl);
1350extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num);
1351extern int ceph_encode_locks_to_buffer(struct inode *inode,
1352				       struct ceph_filelock *flocks,
1353				       int num_fcntl_locks,
1354				       int num_flock_locks);
1355extern int ceph_locks_to_pagelist(struct ceph_filelock *flocks,
1356				  struct ceph_pagelist *pagelist,
1357				  int num_fcntl_locks, int num_flock_locks);
 
1358
1359/* debugfs.c */
1360extern void ceph_fs_debugfs_init(struct ceph_fs_client *client);
1361extern void ceph_fs_debugfs_cleanup(struct ceph_fs_client *client);
1362
1363/* quota.c */
1364
1365enum quota_get_realm {
1366	QUOTA_GET_MAX_FILES,
1367	QUOTA_GET_MAX_BYTES,
1368	QUOTA_GET_ANY
1369};
1370
1371static inline bool __ceph_has_quota(struct ceph_inode_info *ci,
1372				    enum quota_get_realm which)
1373{
1374	bool has_quota = false;
1375
1376	switch (which) {
1377	case QUOTA_GET_MAX_BYTES:
1378		has_quota = !!ci->i_max_bytes;
1379		break;
1380	case QUOTA_GET_MAX_FILES:
1381		has_quota = !!ci->i_max_files;
1382		break;
1383	default:
1384		has_quota = !!(ci->i_max_files || ci->i_max_bytes);
1385	}
1386	return has_quota;
1387}
1388
1389extern void ceph_adjust_quota_realms_count(struct inode *inode, bool inc);
1390
1391static inline void __ceph_update_quota(struct ceph_inode_info *ci,
1392				       u64 max_bytes, u64 max_files)
1393{
1394	bool had_quota, has_quota;
1395	had_quota = __ceph_has_quota(ci, QUOTA_GET_ANY);
1396	ci->i_max_bytes = max_bytes;
1397	ci->i_max_files = max_files;
1398	has_quota = __ceph_has_quota(ci, QUOTA_GET_ANY);
1399
1400	if (had_quota != has_quota)
1401		ceph_adjust_quota_realms_count(&ci->netfs.inode, has_quota);
1402}
1403
1404static inline int __ceph_sparse_read_ext_count(struct inode *inode, u64 len)
1405{
1406	int cnt = 0;
1407
1408	if (IS_ENCRYPTED(inode)) {
1409		cnt = len >> CEPH_FSCRYPT_BLOCK_SHIFT;
1410		if (cnt > CEPH_SPARSE_EXT_ARRAY_INITIAL)
1411			cnt = 0;
1412	}
1413
1414	return cnt;
1415}
1416
1417extern void ceph_handle_quota(struct ceph_mds_client *mdsc,
1418			      struct ceph_mds_session *session,
1419			      struct ceph_msg *msg);
1420extern bool ceph_quota_is_max_files_exceeded(struct inode *inode);
1421extern bool ceph_quota_is_same_realm(struct inode *old, struct inode *new);
1422extern bool ceph_quota_is_max_bytes_exceeded(struct inode *inode,
1423					     loff_t newlen);
1424extern bool ceph_quota_is_max_bytes_approaching(struct inode *inode,
1425						loff_t newlen);
1426extern bool ceph_quota_update_statfs(struct ceph_fs_client *fsc,
1427				     struct kstatfs *buf);
1428extern void ceph_cleanup_quotarealms_inodes(struct ceph_mds_client *mdsc);
1429
1430bool ceph_inc_mds_stopping_blocker(struct ceph_mds_client *mdsc,
1431			       struct ceph_mds_session *session);
1432void ceph_dec_mds_stopping_blocker(struct ceph_mds_client *mdsc);
1433bool ceph_inc_osd_stopping_blocker(struct ceph_mds_client *mdsc);
1434void ceph_dec_osd_stopping_blocker(struct ceph_mds_client *mdsc);
1435#endif /* _FS_CEPH_SUPER_H */