xref: /kernel/linux/linux-6.6/fs/ceph/super.h (revision 62306a36)
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
7#include <asm/unaligned.h>
8#include <linux/backing-dev.h>
9#include <linux/completion.h>
10#include <linux/exportfs.h>
11#include <linux/fs.h>
12#include <linux/mempool.h>
13#include <linux/pagemap.h>
14#include <linux/wait.h>
15#include <linux/writeback.h>
16#include <linux/slab.h>
17#include <linux/posix_acl.h>
18#include <linux/refcount.h>
19#include <linux/security.h>
20#include <linux/netfs.h>
21#include <linux/fscache.h>
22#include <linux/hashtable.h>
23
24#include <linux/ceph/libceph.h>
25#include "crypto.h"
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#define CEPH_4K_BLOCK_SHIFT 12  /* 4 KB */
32
33#define CEPH_MOUNT_OPT_CLEANRECOVER    (1<<1) /* auto reonnect (clean mode) after blocklisted */
34#define CEPH_MOUNT_OPT_DIRSTAT         (1<<4) /* `cat dirname` for stats */
35#define CEPH_MOUNT_OPT_RBYTES          (1<<5) /* dir st_bytes = rbytes */
36#define CEPH_MOUNT_OPT_NOASYNCREADDIR  (1<<7) /* no dcache readdir */
37#define CEPH_MOUNT_OPT_INO32           (1<<8) /* 32 bit inos */
38#define CEPH_MOUNT_OPT_DCACHE          (1<<9) /* use dcache for readdir etc */
39#define CEPH_MOUNT_OPT_FSCACHE         (1<<10) /* use fscache */
40#define CEPH_MOUNT_OPT_NOPOOLPERM      (1<<11) /* no pool permission check */
41#define CEPH_MOUNT_OPT_MOUNTWAIT       (1<<12) /* mount waits if no mds is up */
42#define CEPH_MOUNT_OPT_NOQUOTADF       (1<<13) /* no root dir quota in statfs */
43#define CEPH_MOUNT_OPT_NOCOPYFROM      (1<<14) /* don't use RADOS 'copy-from' op */
44#define CEPH_MOUNT_OPT_ASYNC_DIROPS    (1<<15) /* allow async directory ops */
45#define CEPH_MOUNT_OPT_NOPAGECACHE     (1<<16) /* bypass pagecache altogether */
46#define CEPH_MOUNT_OPT_SPARSEREAD      (1<<17) /* always do sparse reads */
47
48#define CEPH_MOUNT_OPT_DEFAULT			\
49	(CEPH_MOUNT_OPT_DCACHE |		\
50	 CEPH_MOUNT_OPT_NOCOPYFROM |		\
51	 CEPH_MOUNT_OPT_ASYNC_DIROPS)
52
53#define ceph_set_mount_opt(fsc, opt) \
54	(fsc)->mount_options->flags |= CEPH_MOUNT_OPT_##opt
55#define ceph_clear_mount_opt(fsc, opt) \
56	(fsc)->mount_options->flags &= ~CEPH_MOUNT_OPT_##opt
57#define ceph_test_mount_opt(fsc, opt) \
58	(!!((fsc)->mount_options->flags & CEPH_MOUNT_OPT_##opt))
59
60/* max size of osd read request, limited by libceph */
61#define CEPH_MAX_READ_SIZE              CEPH_MSG_MAX_DATA_LEN
62/* osd has a configurable limitaion of max write size.
63 * CEPH_MSG_MAX_DATA_LEN should be small enough. */
64#define CEPH_MAX_WRITE_SIZE		CEPH_MSG_MAX_DATA_LEN
65#define CEPH_RASIZE_DEFAULT             (8192*1024)    /* max readahead */
66#define CEPH_MAX_READDIR_DEFAULT        1024
67#define CEPH_MAX_READDIR_BYTES_DEFAULT  (512*1024)
68#define CEPH_SNAPDIRNAME_DEFAULT        ".snap"
69
70/*
71 * Delay telling the MDS we no longer want caps, in case we reopen
72 * the file.  Delay a minimum amount of time, even if we send a cap
73 * message for some other reason.  Otherwise, take the oppotunity to
74 * update the mds to avoid sending another message later.
75 */
76#define CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT      5  /* cap release delay */
77#define CEPH_CAPS_WANTED_DELAY_MAX_DEFAULT     60  /* cap release delay */
78
79struct ceph_mount_options {
80	unsigned int flags;
81
82	unsigned int wsize;            /* max write size */
83	unsigned int rsize;            /* max read size */
84	unsigned int rasize;           /* max readahead */
85	unsigned int congestion_kb;    /* max writeback in flight */
86	unsigned int caps_wanted_delay_min, caps_wanted_delay_max;
87	int caps_max;
88	unsigned int max_readdir;       /* max readdir result (entries) */
89	unsigned int max_readdir_bytes; /* max readdir result (bytes) */
90
91	bool new_dev_syntax;
92
93	/*
94	 * everything above this point can be memcmp'd; everything below
95	 * is handled in compare_mount_options()
96	 */
97
98	char *snapdir_name;   /* default ".snap" */
99	char *mds_namespace;  /* default NULL */
100	char *server_path;    /* default NULL (means "/") */
101	char *fscache_uniq;   /* default NULL */
102	char *mon_addr;
103	struct fscrypt_dummy_policy dummy_enc_policy;
104};
105
106/* mount state */
107enum {
108	CEPH_MOUNT_MOUNTING,
109	CEPH_MOUNT_MOUNTED,
110	CEPH_MOUNT_UNMOUNTING,
111	CEPH_MOUNT_UNMOUNTED,
112	CEPH_MOUNT_SHUTDOWN,
113	CEPH_MOUNT_RECOVER,
114	CEPH_MOUNT_FENCE_IO,
115};
116
117#define CEPH_ASYNC_CREATE_CONFLICT_BITS 8
118
119struct ceph_fs_client {
120	struct super_block *sb;
121
122	struct list_head metric_wakeup;
123
124	struct ceph_mount_options *mount_options;
125	struct ceph_client *client;
126
127	int mount_state;
128
129	bool blocklisted;
130
131	bool have_copy_from2;
132
133	u32 filp_gen;
134	loff_t max_file_size;
135
136	struct ceph_mds_client *mdsc;
137
138	atomic_long_t writeback_count;
139	bool write_congested;
140
141	struct workqueue_struct *inode_wq;
142	struct workqueue_struct *cap_wq;
143
144	DECLARE_HASHTABLE(async_unlink_conflict, CEPH_ASYNC_CREATE_CONFLICT_BITS);
145	spinlock_t async_unlink_conflict_lock;
146
147#ifdef CONFIG_DEBUG_FS
148	struct dentry *debugfs_dentry_lru, *debugfs_caps;
149	struct dentry *debugfs_congestion_kb;
150	struct dentry *debugfs_bdi;
151	struct dentry *debugfs_mdsc, *debugfs_mdsmap;
152	struct dentry *debugfs_status;
153	struct dentry *debugfs_mds_sessions;
154	struct dentry *debugfs_metrics_dir;
155#endif
156
157#ifdef CONFIG_CEPH_FSCACHE
158	struct fscache_volume *fscache;
159#endif
160#ifdef CONFIG_FS_ENCRYPTION
161	struct fscrypt_dummy_policy fsc_dummy_enc_policy;
162#endif
163};
164
165/*
166 * File i/o capability.  This tracks shared state with the metadata
167 * server that allows us to cache or writeback attributes or to read
168 * and write data.  For any given inode, we should have one or more
169 * capabilities, one issued by each metadata server, and our
170 * cumulative access is the OR of all issued capabilities.
171 *
172 * Each cap is referenced by the inode's i_caps rbtree and by per-mds
173 * session capability lists.
174 */
175struct ceph_cap {
176	struct ceph_inode_info *ci;
177	struct rb_node ci_node;          /* per-ci cap tree */
178	struct ceph_mds_session *session;
179	struct list_head session_caps;   /* per-session caplist */
180	u64 cap_id;       /* unique cap id (mds provided) */
181	union {
182		/* in-use caps */
183		struct {
184			int issued;       /* latest, from the mds */
185			int implemented;  /* implemented superset of
186					     issued (for revocation) */
187			int mds;	  /* mds index for this cap */
188			int mds_wanted;   /* caps wanted from this mds */
189		};
190		/* caps to release */
191		struct {
192			u64 cap_ino;
193			int queue_release;
194		};
195	};
196	u32 seq, issue_seq, mseq;
197	u32 cap_gen;      /* active/stale cycle */
198	unsigned long last_used;
199	struct list_head caps_item;
200};
201
202#define CHECK_CAPS_AUTHONLY   1  /* only check auth cap */
203#define CHECK_CAPS_FLUSH      2  /* flush any dirty caps */
204#define CHECK_CAPS_NOINVAL    4  /* don't invalidate pagecache */
205
206struct ceph_cap_flush {
207	u64 tid;
208	int caps;
209	bool wake; /* wake up flush waiters when finish ? */
210	bool is_capsnap; /* true means capsnap */
211	struct list_head g_list; // global
212	struct list_head i_list; // per inode
213};
214
215/*
216 * Snapped cap state that is pending flush to mds.  When a snapshot occurs,
217 * we first complete any in-process sync writes and writeback any dirty
218 * data before flushing the snapped state (tracked here) back to the MDS.
219 */
220struct ceph_cap_snap {
221	refcount_t nref;
222	struct list_head ci_item;
223
224	struct ceph_cap_flush cap_flush;
225
226	u64 follows;
227	int issued, dirty;
228	struct ceph_snap_context *context;
229
230	umode_t mode;
231	kuid_t uid;
232	kgid_t gid;
233
234	struct ceph_buffer *xattr_blob;
235	u64 xattr_version;
236
237	u64 size;
238	u64 change_attr;
239	struct timespec64 mtime, atime, ctime, btime;
240	u64 time_warp_seq;
241	u64 truncate_size;
242	u32 truncate_seq;
243	int writing;   /* a sync write is still in progress */
244	int dirty_pages;     /* dirty pages awaiting writeback */
245	bool inline_data;
246	bool need_flush;
247};
248
249static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap)
250{
251	if (refcount_dec_and_test(&capsnap->nref)) {
252		if (capsnap->xattr_blob)
253			ceph_buffer_put(capsnap->xattr_blob);
254		kmem_cache_free(ceph_cap_snap_cachep, capsnap);
255	}
256}
257
258/*
259 * The frag tree describes how a directory is fragmented, potentially across
260 * multiple metadata servers.  It is also used to indicate points where
261 * metadata authority is delegated, and whether/where metadata is replicated.
262 *
263 * A _leaf_ frag will be present in the i_fragtree IFF there is
264 * delegation info.  That is, if mds >= 0 || ndist > 0.
265 */
266#define CEPH_MAX_DIRFRAG_REP 4
267
268struct ceph_inode_frag {
269	struct rb_node node;
270
271	/* fragtree state */
272	u32 frag;
273	int split_by;         /* i.e. 2^(split_by) children */
274
275	/* delegation and replication info */
276	int mds;              /* -1 if same authority as parent */
277	int ndist;            /* >0 if replicated */
278	int dist[CEPH_MAX_DIRFRAG_REP];
279};
280
281/*
282 * We cache inode xattrs as an encoded blob until they are first used,
283 * at which point we parse them into an rbtree.
284 */
285struct ceph_inode_xattr {
286	struct rb_node node;
287
288	const char *name;
289	int name_len;
290	const char *val;
291	int val_len;
292	int dirty;
293
294	int should_free_name;
295	int should_free_val;
296};
297
298/*
299 * Ceph dentry state
300 */
301struct ceph_dentry_info {
302	struct dentry *dentry;
303	struct ceph_mds_session *lease_session;
304	struct list_head lease_list;
305	struct hlist_node hnode;
306	unsigned long flags;
307	int lease_shared_gen;
308	u32 lease_gen;
309	u32 lease_seq;
310	unsigned long lease_renew_after, lease_renew_from;
311	unsigned long time;
312	u64 offset;
313};
314
315#define CEPH_DENTRY_REFERENCED		(1 << 0)
316#define CEPH_DENTRY_LEASE_LIST		(1 << 1)
317#define CEPH_DENTRY_SHRINK_LIST		(1 << 2)
318#define CEPH_DENTRY_PRIMARY_LINK	(1 << 3)
319#define CEPH_DENTRY_ASYNC_UNLINK_BIT	(4)
320#define CEPH_DENTRY_ASYNC_UNLINK	(1 << CEPH_DENTRY_ASYNC_UNLINK_BIT)
321#define CEPH_DENTRY_ASYNC_CREATE_BIT	(5)
322#define CEPH_DENTRY_ASYNC_CREATE	(1 << CEPH_DENTRY_ASYNC_CREATE_BIT)
323
324struct ceph_inode_xattrs_info {
325	/*
326	 * (still encoded) xattr blob. we avoid the overhead of parsing
327	 * this until someone actually calls getxattr, etc.
328	 *
329	 * blob->vec.iov_len == 4 implies there are no xattrs; blob ==
330	 * NULL means we don't know.
331	*/
332	struct ceph_buffer *blob, *prealloc_blob;
333
334	struct rb_root index;
335	bool dirty;
336	int count;
337	int names_size;
338	int vals_size;
339	u64 version, index_version;
340};
341
342/*
343 * Ceph inode.
344 */
345struct ceph_inode_info {
346	struct netfs_inode netfs; /* Netfslib context and vfs inode */
347	struct ceph_vino i_vino;   /* ceph ino + snap */
348
349	spinlock_t i_ceph_lock;
350
351	u64 i_version;
352	u64 i_inline_version;
353	u32 i_time_warp_seq;
354
355	unsigned long i_ceph_flags;
356	atomic64_t i_release_count;
357	atomic64_t i_ordered_count;
358	atomic64_t i_complete_seq[2];
359
360	struct ceph_dir_layout i_dir_layout;
361	struct ceph_file_layout i_layout;
362	struct ceph_file_layout i_cached_layout;	// for async creates
363	char *i_symlink;
364
365	/* for dirs */
366	struct timespec64 i_rctime;
367	u64 i_rbytes, i_rfiles, i_rsubdirs, i_rsnaps;
368	u64 i_files, i_subdirs;
369
370	/* quotas */
371	u64 i_max_bytes, i_max_files;
372
373	s32 i_dir_pin;
374
375	struct rb_root i_fragtree;
376	int i_fragtree_nsplits;
377	struct mutex i_fragtree_mutex;
378
379	struct ceph_inode_xattrs_info i_xattrs;
380
381	/* capabilities.  protected _both_ by i_ceph_lock and cap->session's
382	 * s_mutex. */
383	struct rb_root i_caps;           /* cap list */
384	struct ceph_cap *i_auth_cap;     /* authoritative cap, if any */
385	unsigned i_dirty_caps, i_flushing_caps;     /* mask of dirtied fields */
386
387	/*
388	 * Link to the auth cap's session's s_cap_dirty list. s_cap_dirty
389	 * is protected by the mdsc->cap_dirty_lock, but each individual item
390	 * is also protected by the inode's i_ceph_lock. Walking s_cap_dirty
391	 * requires the mdsc->cap_dirty_lock. List presence for an item can
392	 * be tested under the i_ceph_lock. Changing anything requires both.
393	 */
394	struct list_head i_dirty_item;
395
396	/*
397	 * Link to session's s_cap_flushing list. Protected in a similar
398	 * fashion to i_dirty_item, but also by the s_mutex for changes. The
399	 * s_cap_flushing list can be walked while holding either the s_mutex
400	 * or msdc->cap_dirty_lock. List presence can also be checked while
401	 * holding the i_ceph_lock for this inode.
402	 */
403	struct list_head i_flushing_item;
404
405	/* we need to track cap writeback on a per-cap-bit basis, to allow
406	 * overlapping, pipelined cap flushes to the mds.  we can probably
407	 * reduce the tid to 8 bits if we're concerned about inode size. */
408	struct ceph_cap_flush *i_prealloc_cap_flush;
409	struct list_head i_cap_flush_list;
410	wait_queue_head_t i_cap_wq;      /* threads waiting on a capability */
411	unsigned long i_hold_caps_max; /* jiffies */
412	struct list_head i_cap_delay_list;  /* for delayed cap release to mds */
413	struct ceph_cap_reservation i_cap_migration_resv;
414	struct list_head i_cap_snaps;   /* snapped state pending flush to mds */
415	struct ceph_snap_context *i_head_snapc;  /* set if wr_buffer_head > 0 or
416						    dirty|flushing caps */
417	unsigned i_snap_caps;           /* cap bits for snapped files */
418
419	unsigned long i_last_rd;
420	unsigned long i_last_wr;
421	int i_nr_by_mode[CEPH_FILE_MODE_BITS];  /* open file counts */
422
423	struct mutex i_truncate_mutex;
424	u32 i_truncate_seq;        /* last truncate to smaller size */
425	u64 i_truncate_size;       /*  and the size we last truncated down to */
426	int i_truncate_pending;    /*  still need to call vmtruncate */
427	/*
428	 * For none fscrypt case it equals to i_truncate_size or it will
429	 * equals to fscrypt_file_size
430	 */
431	u64 i_truncate_pagecache_size;
432
433	u64 i_max_size;            /* max file size authorized by mds */
434	u64 i_reported_size; /* (max_)size reported to or requested of mds */
435	u64 i_wanted_max_size;     /* offset we'd like to write too */
436	u64 i_requested_max_size;  /* max_size we've requested */
437
438	/* held references to caps */
439	int i_pin_ref;
440	int i_rd_ref, i_rdcache_ref, i_wr_ref, i_wb_ref, i_fx_ref;
441	int i_wrbuffer_ref, i_wrbuffer_ref_head;
442	atomic_t i_filelock_ref;
443	atomic_t i_shared_gen;       /* increment each time we get FILE_SHARED */
444	u32 i_rdcache_gen;      /* incremented each time we get FILE_CACHE. */
445	u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */
446
447	struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */
448	struct list_head i_unsafe_iops;   /* uncommitted mds inode ops */
449	spinlock_t i_unsafe_lock;
450
451	union {
452		struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */
453		struct ceph_snapid_map *i_snapid_map; /* snapid -> dev_t */
454	};
455	struct list_head i_snap_realm_item;
456	struct list_head i_snap_flush_item;
457	struct timespec64 i_btime;
458	struct timespec64 i_snap_btime;
459
460	struct work_struct i_work;
461	unsigned long  i_work_mask;
462
463#ifdef CONFIG_FS_ENCRYPTION
464	u32 fscrypt_auth_len;
465	u32 fscrypt_file_len;
466	u8 *fscrypt_auth;
467	u8 *fscrypt_file;
468#endif
469};
470
471struct ceph_netfs_request_data {
472	int caps;
473
474	/*
475	 * Maximum size of a file readahead request.
476	 * The fadvise could update the bdi's default ra_pages.
477	 */
478	unsigned int file_ra_pages;
479
480	/* Set it if fadvise disables file readahead entirely */
481	bool file_ra_disabled;
482};
483
484static inline struct ceph_inode_info *
485ceph_inode(const struct inode *inode)
486{
487	return container_of(inode, struct ceph_inode_info, netfs.inode);
488}
489
490static inline struct ceph_fs_client *
491ceph_inode_to_client(const struct inode *inode)
492{
493	return (struct ceph_fs_client *)inode->i_sb->s_fs_info;
494}
495
496static inline struct ceph_fs_client *
497ceph_sb_to_client(const struct super_block *sb)
498{
499	return (struct ceph_fs_client *)sb->s_fs_info;
500}
501
502static inline struct ceph_mds_client *
503ceph_sb_to_mdsc(const struct super_block *sb)
504{
505	return (struct ceph_mds_client *)ceph_sb_to_client(sb)->mdsc;
506}
507
508static inline struct ceph_vino
509ceph_vino(const struct inode *inode)
510{
511	return ceph_inode(inode)->i_vino;
512}
513
514static inline u32 ceph_ino_to_ino32(u64 vino)
515{
516	u32 ino = vino & 0xffffffff;
517	ino ^= vino >> 32;
518	if (!ino)
519		ino = 2;
520	return ino;
521}
522
523/*
524 * Inode numbers in cephfs are 64 bits, but inode->i_ino is 32-bits on
525 * some arches. We generally do not use this value inside the ceph driver, but
526 * we do want to set it to something, so that generic vfs code has an
527 * appropriate value for tracepoints and the like.
528 */
529static inline ino_t ceph_vino_to_ino_t(struct ceph_vino vino)
530{
531	if (sizeof(ino_t) == sizeof(u32))
532		return ceph_ino_to_ino32(vino.ino);
533	return (ino_t)vino.ino;
534}
535
536/* for printf-style formatting */
537#define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap
538
539static inline u64 ceph_ino(struct inode *inode)
540{
541	return ceph_inode(inode)->i_vino.ino;
542}
543
544static inline u64 ceph_snap(struct inode *inode)
545{
546	return ceph_inode(inode)->i_vino.snap;
547}
548
549/**
550 * ceph_present_ino - format an inode number for presentation to userland
551 * @sb: superblock where the inode lives
552 * @ino: inode number to (possibly) convert
553 *
554 * If the user mounted with the ino32 option, then the 64-bit value needs
555 * to be converted to something that can fit inside 32 bits. Note that
556 * internal kernel code never uses this value, so this is entirely for
557 * userland consumption.
558 */
559static inline u64 ceph_present_ino(struct super_block *sb, u64 ino)
560{
561	if (unlikely(ceph_test_mount_opt(ceph_sb_to_client(sb), INO32)))
562		return ceph_ino_to_ino32(ino);
563	return ino;
564}
565
566static inline u64 ceph_present_inode(struct inode *inode)
567{
568	return ceph_present_ino(inode->i_sb, ceph_ino(inode));
569}
570
571static inline int ceph_ino_compare(struct inode *inode, void *data)
572{
573	struct ceph_vino *pvino = (struct ceph_vino *)data;
574	struct ceph_inode_info *ci = ceph_inode(inode);
575	return ci->i_vino.ino == pvino->ino &&
576		ci->i_vino.snap == pvino->snap;
577}
578
579/*
580 * The MDS reserves a set of inodes for its own usage. These should never
581 * be accessible by clients, and so the MDS has no reason to ever hand these
582 * out. The range is CEPH_MDS_INO_MDSDIR_OFFSET..CEPH_INO_SYSTEM_BASE.
583 *
584 * These come from src/mds/mdstypes.h in the ceph sources.
585 */
586#define CEPH_MAX_MDS			0x100
587#define CEPH_NUM_STRAY			10
588#define CEPH_MDS_INO_MDSDIR_OFFSET	(1 * CEPH_MAX_MDS)
589#define CEPH_MDS_INO_LOG_OFFSET		(2 * CEPH_MAX_MDS)
590#define CEPH_INO_SYSTEM_BASE		((6*CEPH_MAX_MDS) + (CEPH_MAX_MDS * CEPH_NUM_STRAY))
591
592static inline bool ceph_vino_is_reserved(const struct ceph_vino vino)
593{
594	if (vino.ino >= CEPH_INO_SYSTEM_BASE ||
595	    vino.ino < CEPH_MDS_INO_MDSDIR_OFFSET)
596		return false;
597
598	/* Don't warn on mdsdirs */
599	WARN_RATELIMIT(vino.ino >= CEPH_MDS_INO_LOG_OFFSET,
600			"Attempt to access reserved inode number 0x%llx",
601			vino.ino);
602	return true;
603}
604
605static inline struct inode *ceph_find_inode(struct super_block *sb,
606					    struct ceph_vino vino)
607{
608	if (ceph_vino_is_reserved(vino))
609		return NULL;
610
611	/*
612	 * NB: The hashval will be run through the fs/inode.c hash function
613	 * anyway, so there is no need to squash the inode number down to
614	 * 32-bits first. Just use low-order bits on arches with 32-bit long.
615	 */
616	return ilookup5(sb, (unsigned long)vino.ino, ceph_ino_compare, &vino);
617}
618
619
620/*
621 * Ceph inode.
622 */
623#define CEPH_I_DIR_ORDERED	(1 << 0)  /* dentries in dir are ordered */
624#define CEPH_I_FLUSH		(1 << 2)  /* do not delay flush of dirty metadata */
625#define CEPH_I_POOL_PERM	(1 << 3)  /* pool rd/wr bits are valid */
626#define CEPH_I_POOL_RD		(1 << 4)  /* can read from pool */
627#define CEPH_I_POOL_WR		(1 << 5)  /* can write to pool */
628#define CEPH_I_SEC_INITED	(1 << 6)  /* security initialized */
629#define CEPH_I_KICK_FLUSH	(1 << 7)  /* kick flushing caps */
630#define CEPH_I_FLUSH_SNAPS	(1 << 8)  /* need flush snapss */
631#define CEPH_I_ERROR_WRITE	(1 << 9) /* have seen write errors */
632#define CEPH_I_ERROR_FILELOCK	(1 << 10) /* have seen file lock errors */
633#define CEPH_I_ODIRECT		(1 << 11) /* inode in direct I/O mode */
634#define CEPH_ASYNC_CREATE_BIT	(12)	  /* async create in flight for this */
635#define CEPH_I_ASYNC_CREATE	(1 << CEPH_ASYNC_CREATE_BIT)
636#define CEPH_I_SHUTDOWN		(1 << 13) /* inode is no longer usable */
637#define CEPH_I_ASYNC_CHECK_CAPS	(1 << 14) /* check caps immediately after async
638					     creating finishes */
639
640/*
641 * Masks of ceph inode work.
642 */
643#define CEPH_I_WORK_WRITEBACK		0
644#define CEPH_I_WORK_INVALIDATE_PAGES	1
645#define CEPH_I_WORK_VMTRUNCATE		2
646#define CEPH_I_WORK_CHECK_CAPS		3
647#define CEPH_I_WORK_FLUSH_SNAPS		4
648
649/*
650 * We set the ERROR_WRITE bit when we start seeing write errors on an inode
651 * and then clear it when they start succeeding. Note that we do a lockless
652 * check first, and only take the lock if it looks like it needs to be changed.
653 * The write submission code just takes this as a hint, so we're not too
654 * worried if a few slip through in either direction.
655 */
656static inline void ceph_set_error_write(struct ceph_inode_info *ci)
657{
658	if (!(READ_ONCE(ci->i_ceph_flags) & CEPH_I_ERROR_WRITE)) {
659		spin_lock(&ci->i_ceph_lock);
660		ci->i_ceph_flags |= CEPH_I_ERROR_WRITE;
661		spin_unlock(&ci->i_ceph_lock);
662	}
663}
664
665static inline void ceph_clear_error_write(struct ceph_inode_info *ci)
666{
667	if (READ_ONCE(ci->i_ceph_flags) & CEPH_I_ERROR_WRITE) {
668		spin_lock(&ci->i_ceph_lock);
669		ci->i_ceph_flags &= ~CEPH_I_ERROR_WRITE;
670		spin_unlock(&ci->i_ceph_lock);
671	}
672}
673
674static inline void __ceph_dir_set_complete(struct ceph_inode_info *ci,
675					   long long release_count,
676					   long long ordered_count)
677{
678	/*
679	 * Makes sure operations that setup readdir cache (update page
680	 * cache and i_size) are strongly ordered w.r.t. the following
681	 * atomic64_set() operations.
682	 */
683	smp_mb();
684	atomic64_set(&ci->i_complete_seq[0], release_count);
685	atomic64_set(&ci->i_complete_seq[1], ordered_count);
686}
687
688static inline void __ceph_dir_clear_complete(struct ceph_inode_info *ci)
689{
690	atomic64_inc(&ci->i_release_count);
691}
692
693static inline void __ceph_dir_clear_ordered(struct ceph_inode_info *ci)
694{
695	atomic64_inc(&ci->i_ordered_count);
696}
697
698static inline bool __ceph_dir_is_complete(struct ceph_inode_info *ci)
699{
700	return atomic64_read(&ci->i_complete_seq[0]) ==
701		atomic64_read(&ci->i_release_count);
702}
703
704static inline bool __ceph_dir_is_complete_ordered(struct ceph_inode_info *ci)
705{
706	return  atomic64_read(&ci->i_complete_seq[0]) ==
707		atomic64_read(&ci->i_release_count) &&
708		atomic64_read(&ci->i_complete_seq[1]) ==
709		atomic64_read(&ci->i_ordered_count);
710}
711
712static inline void ceph_dir_clear_complete(struct inode *inode)
713{
714	__ceph_dir_clear_complete(ceph_inode(inode));
715}
716
717static inline void ceph_dir_clear_ordered(struct inode *inode)
718{
719	__ceph_dir_clear_ordered(ceph_inode(inode));
720}
721
722static inline bool ceph_dir_is_complete_ordered(struct inode *inode)
723{
724	bool ret = __ceph_dir_is_complete_ordered(ceph_inode(inode));
725	smp_rmb();
726	return ret;
727}
728
729/* find a specific frag @f */
730extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci,
731						u32 f);
732
733/*
734 * choose fragment for value @v.  copy frag content to pfrag, if leaf
735 * exists
736 */
737extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v,
738			    struct ceph_inode_frag *pfrag,
739			    int *found);
740
741static inline struct ceph_dentry_info *ceph_dentry(const struct dentry *dentry)
742{
743	return (struct ceph_dentry_info *)dentry->d_fsdata;
744}
745
746/*
747 * caps helpers
748 */
749static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci)
750{
751	return !RB_EMPTY_ROOT(&ci->i_caps);
752}
753
754extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented);
755extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t);
756extern int __ceph_caps_issued_mask_metric(struct ceph_inode_info *ci, int mask,
757					  int t);
758extern int __ceph_caps_issued_other(struct ceph_inode_info *ci,
759				    struct ceph_cap *cap);
760
761static inline int ceph_caps_issued(struct ceph_inode_info *ci)
762{
763	int issued;
764	spin_lock(&ci->i_ceph_lock);
765	issued = __ceph_caps_issued(ci, NULL);
766	spin_unlock(&ci->i_ceph_lock);
767	return issued;
768}
769
770static inline int ceph_caps_issued_mask_metric(struct ceph_inode_info *ci,
771					       int mask, int touch)
772{
773	int r;
774	spin_lock(&ci->i_ceph_lock);
775	r = __ceph_caps_issued_mask_metric(ci, mask, touch);
776	spin_unlock(&ci->i_ceph_lock);
777	return r;
778}
779
780static inline int __ceph_caps_dirty(struct ceph_inode_info *ci)
781{
782	return ci->i_dirty_caps | ci->i_flushing_caps;
783}
784extern struct ceph_cap_flush *ceph_alloc_cap_flush(void);
785extern void ceph_free_cap_flush(struct ceph_cap_flush *cf);
786extern int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask,
787				  struct ceph_cap_flush **pcf);
788
789extern int __ceph_caps_revoking_other(struct ceph_inode_info *ci,
790				      struct ceph_cap *ocap, int mask);
791extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask);
792extern int __ceph_caps_used(struct ceph_inode_info *ci);
793
794static inline bool __ceph_is_file_opened(struct ceph_inode_info *ci)
795{
796	return ci->i_nr_by_mode[0];
797}
798extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci);
799extern int __ceph_caps_wanted(struct ceph_inode_info *ci);
800
801/* what the mds thinks we want */
802extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci, bool check);
803
804extern void ceph_caps_init(struct ceph_mds_client *mdsc);
805extern void ceph_caps_finalize(struct ceph_mds_client *mdsc);
806extern void ceph_adjust_caps_max_min(struct ceph_mds_client *mdsc,
807				     struct ceph_mount_options *fsopt);
808extern int ceph_reserve_caps(struct ceph_mds_client *mdsc,
809			     struct ceph_cap_reservation *ctx, int need);
810extern void ceph_unreserve_caps(struct ceph_mds_client *mdsc,
811			       struct ceph_cap_reservation *ctx);
812extern void ceph_reservation_status(struct ceph_fs_client *client,
813				    int *total, int *avail, int *used,
814				    int *reserved, int *min);
815extern void change_auth_cap_ses(struct ceph_inode_info *ci,
816				struct ceph_mds_session *session);
817
818
819
820/*
821 * we keep buffered readdir results attached to file->private_data
822 */
823#define CEPH_F_SYNC     1
824#define CEPH_F_ATEND    2
825
826struct ceph_file_info {
827	short fmode;     /* initialized on open */
828	short flags;     /* CEPH_F_* */
829
830	spinlock_t rw_contexts_lock;
831	struct list_head rw_contexts;
832
833	u32 filp_gen;
834};
835
836struct ceph_dir_file_info {
837	struct ceph_file_info file_info;
838
839	/* readdir: position within the dir */
840	u32 frag;
841	struct ceph_mds_request *last_readdir;
842
843	/* readdir: position within a frag */
844	unsigned next_offset;  /* offset of next chunk (last_name's + 1) */
845	char *last_name;       /* last entry in previous chunk */
846	long long dir_release_count;
847	long long dir_ordered_count;
848	int readdir_cache_idx;
849
850	/* used for -o dirstat read() on directory thing */
851	char *dir_info;
852	int dir_info_len;
853};
854
855struct ceph_rw_context {
856	struct list_head list;
857	struct task_struct *thread;
858	int caps;
859};
860
861#define CEPH_DEFINE_RW_CONTEXT(_name, _caps)	\
862	struct ceph_rw_context _name = {	\
863		.thread = current,		\
864		.caps = _caps,			\
865	}
866
867static inline void ceph_add_rw_context(struct ceph_file_info *cf,
868				       struct ceph_rw_context *ctx)
869{
870	spin_lock(&cf->rw_contexts_lock);
871	list_add(&ctx->list, &cf->rw_contexts);
872	spin_unlock(&cf->rw_contexts_lock);
873}
874
875static inline void ceph_del_rw_context(struct ceph_file_info *cf,
876				       struct ceph_rw_context *ctx)
877{
878	spin_lock(&cf->rw_contexts_lock);
879	list_del(&ctx->list);
880	spin_unlock(&cf->rw_contexts_lock);
881}
882
883static inline struct ceph_rw_context*
884ceph_find_rw_context(struct ceph_file_info *cf)
885{
886	struct ceph_rw_context *ctx, *found = NULL;
887	spin_lock(&cf->rw_contexts_lock);
888	list_for_each_entry(ctx, &cf->rw_contexts, list) {
889		if (ctx->thread == current) {
890			found = ctx;
891			break;
892		}
893	}
894	spin_unlock(&cf->rw_contexts_lock);
895	return found;
896}
897
898struct ceph_readdir_cache_control {
899	struct page  *page;
900	struct dentry **dentries;
901	int index;
902};
903
904/*
905 * A "snap realm" describes a subset of the file hierarchy sharing
906 * the same set of snapshots that apply to it.  The realms themselves
907 * are organized into a hierarchy, such that children inherit (some of)
908 * the snapshots of their parents.
909 *
910 * All inodes within the realm that have capabilities are linked into a
911 * per-realm list.
912 */
913struct ceph_snap_realm {
914	u64 ino;
915	struct inode *inode;
916	atomic_t nref;
917	struct rb_node node;
918
919	u64 created, seq;
920	u64 parent_ino;
921	u64 parent_since;   /* snapid when our current parent became so */
922
923	u64 *prior_parent_snaps;      /* snaps inherited from any parents we */
924	u32 num_prior_parent_snaps;   /*  had prior to parent_since */
925	u64 *snaps;                   /* snaps specific to this realm */
926	u32 num_snaps;
927
928	struct ceph_snap_realm *parent;
929	struct list_head children;       /* list of child realms */
930	struct list_head child_item;
931
932	struct list_head empty_item;     /* if i have ref==0 */
933
934	struct list_head dirty_item;     /* if realm needs new context */
935
936	struct list_head rebuild_item;   /* rebuild snap realms _downward_ in hierarchy */
937
938	/* the current set of snaps for this realm */
939	struct ceph_snap_context *cached_context;
940
941	struct list_head inodes_with_caps;
942	spinlock_t inodes_with_caps_lock;
943};
944
945static inline int default_congestion_kb(void)
946{
947	int congestion_kb;
948
949	/*
950	 * Copied from NFS
951	 *
952	 * congestion size, scale with available memory.
953	 *
954	 *  64MB:    8192k
955	 * 128MB:   11585k
956	 * 256MB:   16384k
957	 * 512MB:   23170k
958	 *   1GB:   32768k
959	 *   2GB:   46340k
960	 *   4GB:   65536k
961	 *   8GB:   92681k
962	 *  16GB:  131072k
963	 *
964	 * This allows larger machines to have larger/more transfers.
965	 * Limit the default to 256M
966	 */
967	congestion_kb = (16*int_sqrt(totalram_pages())) << (PAGE_SHIFT-10);
968	if (congestion_kb > 256*1024)
969		congestion_kb = 256*1024;
970
971	return congestion_kb;
972}
973
974
975/* super.c */
976extern int ceph_force_reconnect(struct super_block *sb);
977/* snap.c */
978struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc,
979					       u64 ino);
980extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc,
981				struct ceph_snap_realm *realm);
982extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc,
983				struct ceph_snap_realm *realm);
984extern int ceph_update_snap_trace(struct ceph_mds_client *m,
985				  void *p, void *e, bool deletion,
986				  struct ceph_snap_realm **realm_ret);
987void ceph_change_snap_realm(struct inode *inode, struct ceph_snap_realm *realm);
988extern void ceph_handle_snap(struct ceph_mds_client *mdsc,
989			     struct ceph_mds_session *session,
990			     struct ceph_msg *msg);
991extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci,
992				  struct ceph_cap_snap *capsnap);
993extern void ceph_cleanup_global_and_empty_realms(struct ceph_mds_client *mdsc);
994
995extern struct ceph_snapid_map *ceph_get_snapid_map(struct ceph_mds_client *mdsc,
996						   u64 snap);
997extern void ceph_put_snapid_map(struct ceph_mds_client* mdsc,
998				struct ceph_snapid_map *sm);
999extern void ceph_trim_snapid_map(struct ceph_mds_client *mdsc);
1000extern void ceph_cleanup_snapid_map(struct ceph_mds_client *mdsc);
1001void ceph_umount_begin(struct super_block *sb);
1002
1003
1004/*
1005 * a cap_snap is "pending" if it is still awaiting an in-progress
1006 * sync write (that may/may not still update size, mtime, etc.).
1007 */
1008static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci)
1009{
1010	return !list_empty(&ci->i_cap_snaps) &&
1011	       list_last_entry(&ci->i_cap_snaps, struct ceph_cap_snap,
1012			       ci_item)->writing;
1013}
1014
1015/* inode.c */
1016struct ceph_mds_reply_info_in;
1017struct ceph_mds_reply_dirfrag;
1018struct ceph_acl_sec_ctx;
1019
1020extern const struct inode_operations ceph_file_iops;
1021
1022extern struct inode *ceph_alloc_inode(struct super_block *sb);
1023extern void ceph_evict_inode(struct inode *inode);
1024extern void ceph_free_inode(struct inode *inode);
1025
1026struct inode *ceph_new_inode(struct inode *dir, struct dentry *dentry,
1027			     umode_t *mode, struct ceph_acl_sec_ctx *as_ctx);
1028void ceph_as_ctx_to_req(struct ceph_mds_request *req,
1029			struct ceph_acl_sec_ctx *as_ctx);
1030
1031extern struct inode *ceph_get_inode(struct super_block *sb,
1032				    struct ceph_vino vino,
1033				    struct inode *newino);
1034extern struct inode *ceph_get_snapdir(struct inode *parent);
1035extern int ceph_fill_file_size(struct inode *inode, int issued,
1036			       u32 truncate_seq, u64 truncate_size, u64 size);
1037extern void ceph_fill_file_time(struct inode *inode, int issued,
1038				u64 time_warp_seq, struct timespec64 *ctime,
1039				struct timespec64 *mtime,
1040				struct timespec64 *atime);
1041extern int ceph_fill_inode(struct inode *inode, struct page *locked_page,
1042		    struct ceph_mds_reply_info_in *iinfo,
1043		    struct ceph_mds_reply_dirfrag *dirinfo,
1044		    struct ceph_mds_session *session, int cap_fmode,
1045		    struct ceph_cap_reservation *caps_reservation);
1046extern int ceph_fill_trace(struct super_block *sb,
1047			   struct ceph_mds_request *req);
1048extern int ceph_readdir_prepopulate(struct ceph_mds_request *req,
1049				    struct ceph_mds_session *session);
1050
1051extern int ceph_inode_holds_cap(struct inode *inode, int mask);
1052
1053extern bool ceph_inode_set_size(struct inode *inode, loff_t size);
1054extern void __ceph_do_pending_vmtruncate(struct inode *inode);
1055
1056void ceph_queue_inode_work(struct inode *inode, int work_bit);
1057
1058static inline void ceph_queue_vmtruncate(struct inode *inode)
1059{
1060	ceph_queue_inode_work(inode, CEPH_I_WORK_VMTRUNCATE);
1061}
1062
1063static inline void ceph_queue_invalidate(struct inode *inode)
1064{
1065	ceph_queue_inode_work(inode, CEPH_I_WORK_INVALIDATE_PAGES);
1066}
1067
1068static inline void ceph_queue_writeback(struct inode *inode)
1069{
1070	ceph_queue_inode_work(inode, CEPH_I_WORK_WRITEBACK);
1071}
1072
1073static inline void ceph_queue_check_caps(struct inode *inode)
1074{
1075	ceph_queue_inode_work(inode, CEPH_I_WORK_CHECK_CAPS);
1076}
1077
1078static inline void ceph_queue_flush_snaps(struct inode *inode)
1079{
1080	ceph_queue_inode_work(inode, CEPH_I_WORK_FLUSH_SNAPS);
1081}
1082
1083extern int ceph_try_to_choose_auth_mds(struct inode *inode, int mask);
1084extern int __ceph_do_getattr(struct inode *inode, struct page *locked_page,
1085			     int mask, bool force);
1086static inline int ceph_do_getattr(struct inode *inode, int mask, bool force)
1087{
1088	return __ceph_do_getattr(inode, NULL, mask, force);
1089}
1090extern int ceph_permission(struct mnt_idmap *idmap,
1091			   struct inode *inode, int mask);
1092
1093struct ceph_iattr {
1094	struct ceph_fscrypt_auth	*fscrypt_auth;
1095};
1096
1097extern int __ceph_setattr(struct inode *inode, struct iattr *attr,
1098			  struct ceph_iattr *cia);
1099extern int ceph_setattr(struct mnt_idmap *idmap,
1100			struct dentry *dentry, struct iattr *attr);
1101extern int ceph_getattr(struct mnt_idmap *idmap,
1102			const struct path *path, struct kstat *stat,
1103			u32 request_mask, unsigned int flags);
1104void ceph_inode_shutdown(struct inode *inode);
1105
1106static inline bool ceph_inode_is_shutdown(struct inode *inode)
1107{
1108	unsigned long flags = READ_ONCE(ceph_inode(inode)->i_ceph_flags);
1109	struct ceph_fs_client *fsc = ceph_inode_to_client(inode);
1110	int state = READ_ONCE(fsc->mount_state);
1111
1112	return (flags & CEPH_I_SHUTDOWN) || state >= CEPH_MOUNT_SHUTDOWN;
1113}
1114
1115/* xattr.c */
1116int __ceph_setxattr(struct inode *, const char *, const void *, size_t, int);
1117int ceph_do_getvxattr(struct inode *inode, const char *name, void *value, size_t size);
1118ssize_t __ceph_getxattr(struct inode *, const char *, void *, size_t);
1119extern ssize_t ceph_listxattr(struct dentry *, char *, size_t);
1120extern struct ceph_buffer *__ceph_build_xattrs_blob(struct ceph_inode_info *ci);
1121extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci);
1122extern const struct xattr_handler *ceph_xattr_handlers[];
1123
1124struct ceph_acl_sec_ctx {
1125#ifdef CONFIG_CEPH_FS_POSIX_ACL
1126	void *default_acl;
1127	void *acl;
1128#endif
1129#ifdef CONFIG_CEPH_FS_SECURITY_LABEL
1130	void *sec_ctx;
1131	u32 sec_ctxlen;
1132#endif
1133#ifdef CONFIG_FS_ENCRYPTION
1134	struct ceph_fscrypt_auth *fscrypt_auth;
1135#endif
1136	struct ceph_pagelist *pagelist;
1137};
1138
1139#ifdef CONFIG_SECURITY
1140extern bool ceph_security_xattr_deadlock(struct inode *in);
1141extern bool ceph_security_xattr_wanted(struct inode *in);
1142#else
1143static inline bool ceph_security_xattr_deadlock(struct inode *in)
1144{
1145	return false;
1146}
1147static inline bool ceph_security_xattr_wanted(struct inode *in)
1148{
1149	return false;
1150}
1151#endif
1152
1153#ifdef CONFIG_CEPH_FS_SECURITY_LABEL
1154extern int ceph_security_init_secctx(struct dentry *dentry, umode_t mode,
1155				     struct ceph_acl_sec_ctx *ctx);
1156static inline void ceph_security_invalidate_secctx(struct inode *inode)
1157{
1158	security_inode_invalidate_secctx(inode);
1159}
1160#else
1161static inline int ceph_security_init_secctx(struct dentry *dentry, umode_t mode,
1162					    struct ceph_acl_sec_ctx *ctx)
1163{
1164	return 0;
1165}
1166static inline void ceph_security_invalidate_secctx(struct inode *inode)
1167{
1168}
1169#endif
1170
1171void ceph_release_acl_sec_ctx(struct ceph_acl_sec_ctx *as_ctx);
1172
1173/* acl.c */
1174#ifdef CONFIG_CEPH_FS_POSIX_ACL
1175
1176struct posix_acl *ceph_get_acl(struct inode *, int, bool);
1177int ceph_set_acl(struct mnt_idmap *idmap,
1178		 struct dentry *dentry, struct posix_acl *acl, int type);
1179int ceph_pre_init_acls(struct inode *dir, umode_t *mode,
1180		       struct ceph_acl_sec_ctx *as_ctx);
1181void ceph_init_inode_acls(struct inode *inode,
1182			  struct ceph_acl_sec_ctx *as_ctx);
1183
1184static inline void ceph_forget_all_cached_acls(struct inode *inode)
1185{
1186       forget_all_cached_acls(inode);
1187}
1188
1189#else
1190
1191#define ceph_get_acl NULL
1192#define ceph_set_acl NULL
1193
1194static inline int ceph_pre_init_acls(struct inode *dir, umode_t *mode,
1195				     struct ceph_acl_sec_ctx *as_ctx)
1196{
1197	return 0;
1198}
1199static inline void ceph_init_inode_acls(struct inode *inode,
1200					struct ceph_acl_sec_ctx *as_ctx)
1201{
1202}
1203static inline int ceph_acl_chmod(struct dentry *dentry, struct inode *inode)
1204{
1205	return 0;
1206}
1207
1208static inline void ceph_forget_all_cached_acls(struct inode *inode)
1209{
1210}
1211
1212#endif
1213
1214/* caps.c */
1215extern const char *ceph_cap_string(int c);
1216extern void ceph_handle_caps(struct ceph_mds_session *session,
1217			     struct ceph_msg *msg);
1218extern struct ceph_cap *ceph_get_cap(struct ceph_mds_client *mdsc,
1219				     struct ceph_cap_reservation *ctx);
1220extern void ceph_add_cap(struct inode *inode,
1221			 struct ceph_mds_session *session, u64 cap_id,
1222			 unsigned issued, unsigned wanted,
1223			 unsigned cap, unsigned seq, u64 realmino, int flags,
1224			 struct ceph_cap **new_cap);
1225extern void __ceph_remove_cap(struct ceph_cap *cap, bool queue_release);
1226extern void ceph_remove_cap(struct ceph_cap *cap, bool queue_release);
1227extern void __ceph_remove_caps(struct ceph_inode_info *ci);
1228extern void ceph_put_cap(struct ceph_mds_client *mdsc,
1229			 struct ceph_cap *cap);
1230extern int ceph_is_any_caps(struct inode *inode);
1231
1232extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc);
1233extern int ceph_fsync(struct file *file, loff_t start, loff_t end,
1234		      int datasync);
1235extern void ceph_early_kick_flushing_caps(struct ceph_mds_client *mdsc,
1236					  struct ceph_mds_session *session);
1237extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc,
1238				    struct ceph_mds_session *session);
1239void ceph_kick_flushing_inode_caps(struct ceph_mds_session *session,
1240				   struct ceph_inode_info *ci);
1241extern struct ceph_cap *__get_cap_for_mds(struct ceph_inode_info *ci,
1242					  int mds);
1243extern struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci,
1244					     int mds);
1245extern void ceph_take_cap_refs(struct ceph_inode_info *ci, int caps,
1246				bool snap_rwsem_locked);
1247extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps);
1248extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had);
1249extern void ceph_put_cap_refs_async(struct ceph_inode_info *ci, int had);
1250extern void ceph_put_cap_refs_no_check_caps(struct ceph_inode_info *ci,
1251					    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 int  ceph_drop_caps_for_unlink(struct inode *inode);
1267extern int ceph_encode_inode_release(void **p, struct inode *inode,
1268				     int mds, int drop, int unless, int force);
1269extern int ceph_encode_dentry_release(void **p, struct dentry *dn,
1270				      struct inode *dir,
1271				      int mds, int drop, int unless);
1272
1273extern int __ceph_get_caps(struct inode *inode, struct ceph_file_info *fi,
1274			   int need, int want, loff_t endoff, int *got);
1275extern int ceph_get_caps(struct file *filp, int need, int want,
1276			 loff_t endoff, int *got);
1277extern int ceph_try_get_caps(struct inode *inode,
1278			     int need, int want, bool nonblock, int *got);
1279
1280/* for counting open files by mode */
1281extern void ceph_get_fmode(struct ceph_inode_info *ci, int mode, int count);
1282extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode, int count);
1283extern void __ceph_touch_fmode(struct ceph_inode_info *ci,
1284			       struct ceph_mds_client *mdsc, int fmode);
1285
1286/* addr.c */
1287extern const struct address_space_operations ceph_aops;
1288extern const struct netfs_request_ops ceph_netfs_ops;
1289extern int ceph_mmap(struct file *file, struct vm_area_struct *vma);
1290extern int ceph_uninline_data(struct file *file);
1291extern int ceph_pool_perm_check(struct inode *inode, int need);
1292extern void ceph_pool_perm_destroy(struct ceph_mds_client* mdsc);
1293int ceph_purge_inode_cap(struct inode *inode, struct ceph_cap *cap, bool *invalidate);
1294
1295static inline bool ceph_has_inline_data(struct ceph_inode_info *ci)
1296{
1297	if (ci->i_inline_version == CEPH_INLINE_NONE ||
1298	    ci->i_inline_version == 1) /* initial version, no data */
1299		return false;
1300	return true;
1301}
1302
1303/* file.c */
1304extern const struct file_operations ceph_file_fops;
1305
1306extern int ceph_renew_caps(struct inode *inode, int fmode);
1307extern int ceph_open(struct inode *inode, struct file *file);
1308extern int ceph_atomic_open(struct inode *dir, struct dentry *dentry,
1309			    struct file *file, unsigned flags, umode_t mode);
1310extern ssize_t __ceph_sync_read(struct inode *inode, loff_t *ki_pos,
1311				struct iov_iter *to, int *retry_op,
1312				u64 *last_objver);
1313extern int ceph_release(struct inode *inode, struct file *filp);
1314extern void ceph_fill_inline_data(struct inode *inode, struct page *locked_page,
1315				  char *data, size_t len);
1316
1317/* dir.c */
1318extern const struct file_operations ceph_dir_fops;
1319extern const struct file_operations ceph_snapdir_fops;
1320extern const struct inode_operations ceph_dir_iops;
1321extern const struct inode_operations ceph_snapdir_iops;
1322extern const struct dentry_operations ceph_dentry_ops;
1323
1324extern loff_t ceph_make_fpos(unsigned high, unsigned off, bool hash_order);
1325extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry);
1326extern struct dentry *ceph_handle_snapdir(struct ceph_mds_request *req,
1327			       struct dentry *dentry);
1328extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req,
1329					 struct dentry *dentry, int err);
1330
1331extern void __ceph_dentry_lease_touch(struct ceph_dentry_info *di);
1332extern void __ceph_dentry_dir_lease_touch(struct ceph_dentry_info *di);
1333extern void ceph_invalidate_dentry_lease(struct dentry *dentry);
1334extern int ceph_trim_dentries(struct ceph_mds_client *mdsc);
1335extern unsigned ceph_dentry_hash(struct inode *dir, struct dentry *dn);
1336extern void ceph_readdir_cache_release(struct ceph_readdir_cache_control *ctl);
1337
1338/* ioctl.c */
1339extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
1340
1341/* export.c */
1342extern const struct export_operations ceph_export_ops;
1343struct inode *ceph_lookup_inode(struct super_block *sb, u64 ino);
1344
1345/* locks.c */
1346extern __init void ceph_flock_init(void);
1347extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl);
1348extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl);
1349extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num);
1350extern int ceph_encode_locks_to_buffer(struct inode *inode,
1351				       struct ceph_filelock *flocks,
1352				       int num_fcntl_locks,
1353				       int num_flock_locks);
1354extern int ceph_locks_to_pagelist(struct ceph_filelock *flocks,
1355				  struct ceph_pagelist *pagelist,
1356				  int num_fcntl_locks, int num_flock_locks);
1357
1358/* debugfs.c */
1359extern void ceph_fs_debugfs_init(struct ceph_fs_client *client);
1360extern void ceph_fs_debugfs_cleanup(struct ceph_fs_client *client);
1361
1362/* quota.c */
1363
1364enum quota_get_realm {
1365	QUOTA_GET_MAX_FILES,
1366	QUOTA_GET_MAX_BYTES,
1367	QUOTA_GET_ANY
1368};
1369
1370static inline bool __ceph_has_quota(struct ceph_inode_info *ci,
1371				    enum quota_get_realm which)
1372{
1373	bool has_quota = false;
1374
1375	switch (which) {
1376	case QUOTA_GET_MAX_BYTES:
1377		has_quota = !!ci->i_max_bytes;
1378		break;
1379	case QUOTA_GET_MAX_FILES:
1380		has_quota = !!ci->i_max_files;
1381		break;
1382	default:
1383		has_quota = !!(ci->i_max_files || ci->i_max_bytes);
1384	}
1385	return has_quota;
1386}
1387
1388extern void ceph_adjust_quota_realms_count(struct inode *inode, bool inc);
1389
1390static inline void __ceph_update_quota(struct ceph_inode_info *ci,
1391				       u64 max_bytes, u64 max_files)
1392{
1393	bool had_quota, has_quota;
1394	had_quota = __ceph_has_quota(ci, QUOTA_GET_ANY);
1395	ci->i_max_bytes = max_bytes;
1396	ci->i_max_files = max_files;
1397	has_quota = __ceph_has_quota(ci, QUOTA_GET_ANY);
1398
1399	if (had_quota != has_quota)
1400		ceph_adjust_quota_realms_count(&ci->netfs.inode, has_quota);
1401}
1402
1403extern void ceph_handle_quota(struct ceph_mds_client *mdsc,
1404			      struct ceph_mds_session *session,
1405			      struct ceph_msg *msg);
1406extern bool ceph_quota_is_max_files_exceeded(struct inode *inode);
1407extern bool ceph_quota_is_same_realm(struct inode *old, struct inode *new);
1408extern bool ceph_quota_is_max_bytes_exceeded(struct inode *inode,
1409					     loff_t newlen);
1410extern bool ceph_quota_is_max_bytes_approaching(struct inode *inode,
1411						loff_t newlen);
1412extern bool ceph_quota_update_statfs(struct ceph_fs_client *fsc,
1413				     struct kstatfs *buf);
1414extern void ceph_cleanup_quotarealms_inodes(struct ceph_mds_client *mdsc);
1415
1416bool ceph_inc_mds_stopping_blocker(struct ceph_mds_client *mdsc,
1417			       struct ceph_mds_session *session);
1418void ceph_dec_mds_stopping_blocker(struct ceph_mds_client *mdsc);
1419bool ceph_inc_osd_stopping_blocker(struct ceph_mds_client *mdsc);
1420void ceph_dec_osd_stopping_blocker(struct ceph_mds_client *mdsc);
1421#endif /* _FS_CEPH_SUPER_H */
1422