forked from ~ljy/RK356X_SDK_RELEASE
blame | history | patch | commit | commitdiff
hc
2023-12-11 072de836f53be56a70cecf70b43ae43b7ce17376 
 kernel/fs/ceph/mds_client.h
....@@ -10,23 +10,39 @@
1010 #include <linux/spinlock.h>
1111 #include <linux/refcount.h>
1212 #include <linux/utsname.h>
13
+#include <linux/ktime.h>
1314 
1415 #include <linux/ceph/types.h>
1516 #include <linux/ceph/messenger.h>
1617 #include <linux/ceph/mdsmap.h>
1718 #include <linux/ceph/auth.h>
1819 
20
+#include "metric.h"
21
+#include "super.h"
22
+
1923 /* The first 8 bits are reserved for old ceph releases */
20
-#define CEPHFS_FEATURE_MIMIC    8
24
+enum ceph_feature_type {
25
+	CEPHFS_FEATURE_MIMIC = 8,
26
+	CEPHFS_FEATURE_REPLY_ENCODING,
27
+	CEPHFS_FEATURE_RECLAIM_CLIENT,
28
+	CEPHFS_FEATURE_LAZY_CAP_WANTED,
29
+	CEPHFS_FEATURE_MULTI_RECONNECT,
30
+	CEPHFS_FEATURE_DELEG_INO,
31
+	CEPHFS_FEATURE_METRIC_COLLECT,
2132 
22
-#define CEPHFS_FEATURES_ALL {           \
23
-  0, 1, 2, 3, 4, 5, 6, 7,		\
24
-  CEPHFS_FEATURE_MIMIC,                 \
33
+	CEPHFS_FEATURE_MAX = CEPHFS_FEATURE_METRIC_COLLECT,
34
+};
35
+
36
+#define CEPHFS_FEATURES_CLIENT_SUPPORTED {	\
37
+	0, 1, 2, 3, 4, 5, 6, 7,			\
38
+	CEPHFS_FEATURE_MIMIC,			\
39
+	CEPHFS_FEATURE_REPLY_ENCODING,		\
40
+	CEPHFS_FEATURE_LAZY_CAP_WANTED,		\
41
+	CEPHFS_FEATURE_MULTI_RECONNECT,		\
42
+	CEPHFS_FEATURE_DELEG_INO,		\
43
+	CEPHFS_FEATURE_METRIC_COLLECT,		\
2544 }
26
-
27
-#define CEPHFS_FEATURES_CLIENT_SUPPORTED CEPHFS_FEATURES_ALL
2845 #define CEPHFS_FEATURES_CLIENT_REQUIRED {}
29
-
3046 
3147 /*
3248  * Some lock dependencies:
....@@ -63,6 +79,10 @@
6379 	char *pool_ns_data;
6480 	u64 max_bytes;
6581 	u64 max_files;
82
+	s32 dir_pin;
83
+	struct ceph_timespec btime;
84
+	struct ceph_timespec snap_btime;
85
+	u64 change_attr;
6686 };
6787 
6888 struct ceph_mds_reply_dir_entry {
....@@ -139,10 +159,11 @@
139159 	CEPH_MDS_SESSION_OPENING = 2,
140160 	CEPH_MDS_SESSION_OPEN = 3,
141161 	CEPH_MDS_SESSION_HUNG = 4,
142
-	CEPH_MDS_SESSION_CLOSING = 5,
143
-	CEPH_MDS_SESSION_RESTARTING = 6,
144
-	CEPH_MDS_SESSION_RECONNECTING = 7,
145
-	CEPH_MDS_SESSION_REJECTED = 8,
162
+	CEPH_MDS_SESSION_RESTARTING = 5,
163
+	CEPH_MDS_SESSION_RECONNECTING = 6,
164
+	CEPH_MDS_SESSION_CLOSING = 7,
165
+	CEPH_MDS_SESSION_CLOSED = 8,
166
+	CEPH_MDS_SESSION_REJECTED = 9,
146167 };
147168 
148169 struct ceph_mds_session {
....@@ -150,6 +171,7 @@
150171 	int               s_mds;
151172 	int               s_state;
152173 	unsigned long     s_ttl;      /* time until mds kills us */
174
+	unsigned long	  s_features;
153175 	u64               s_seq;      /* incoming msg seq # */
154176 	struct mutex      s_mutex;    /* serialize session messages */
155177 
....@@ -164,22 +186,28 @@
164186 
165187 	/* protected by s_cap_lock */
166188 	spinlock_t        s_cap_lock;
189
+	refcount_t        s_ref;
167190 	struct list_head  s_caps;     /* all caps issued by this session */
168
-	int               s_nr_caps, s_trim_caps;
191
+	struct ceph_cap  *s_cap_iterator;
192
+	int               s_nr_caps;
169193 	int               s_num_cap_releases;
170194 	int		  s_cap_reconnect;
171195 	int		  s_readonly;
172196 	struct list_head  s_cap_releases; /* waiting cap_release messages */
173
-	struct ceph_cap  *s_cap_iterator;
197
+	struct work_struct s_cap_release_work;
174198 
175
-	/* protected by mutex */
199
+	/* See ceph_inode_info->i_dirty_item. */
200
+	struct list_head  s_cap_dirty;	      /* inodes w/ dirty caps */
201
+
202
+	/* See ceph_inode_info->i_flushing_item. */
176203 	struct list_head  s_cap_flushing;     /* inodes w/ flushing caps */
204
+
177205 	unsigned long     s_renew_requested; /* last time we sent a renew req */
178206 	u64               s_renew_seq;
179207 
180
-	refcount_t          s_ref;
181208 	struct list_head  s_waiting;  /* waiting requests */
182209 	struct list_head  s_unsafe;   /* unsafe requests */
210
+	struct xarray	  s_delegated_inos;
183211 };
184212 
185213 /*
....@@ -213,6 +241,7 @@
213241 	struct rb_node r_node;
214242 	struct ceph_mds_client *r_mdsc;
215243 
244
+	struct kref       r_kref;
216245 	int r_op;                    /* mds op code */
217246 
218247 	/* operation on what? */
....@@ -233,6 +262,7 @@
233262 #define CEPH_MDS_R_GOT_RESULT		(5) /* got a result */
234263 #define CEPH_MDS_R_DID_PREPOPULATE	(6) /* prepopulated readdir */
235264 #define CEPH_MDS_R_PARENT_LOCKED	(7) /* is r_parent->i_rwsem wlocked? */
265
+#define CEPH_MDS_R_ASYNC		(8) /* async request */
236266 	unsigned long	r_req_flags;
237267 
238268 	struct mutex r_fill_mutex;
....@@ -241,6 +271,7 @@
241271 	int r_fmode;        /* file mode, if expecting cap */
242272 	kuid_t r_uid;
243273 	kgid_t r_gid;
274
+	int r_request_release_offset;
244275 	struct timespec64 r_stamp;
245276 
246277 	/* for choosing which mds to send this request to */
....@@ -258,14 +289,20 @@
258289 	int r_old_inode_drop, r_old_inode_unless;
259290 
260291 	struct ceph_msg  *r_request;  /* original request */
261
-	int r_request_release_offset;
262292 	struct ceph_msg  *r_reply;
263293 	struct ceph_mds_reply_info_parsed r_reply_info;
264
-	struct page *r_locked_page;
265294 	int r_err;
295
+
296
+
297
+	struct page *r_locked_page;
298
+	int r_dir_caps;
299
+	int r_num_caps;
300
+	u32               r_readdir_offset;
266301 
267302 	unsigned long r_timeout;  /* optional.  jiffies, 0 is "wait forever" */
268303 	unsigned long r_started;  /* start time to measure timeout against */
304
+	unsigned long r_start_latency;  /* start time to measure latency */
305
+	unsigned long r_end_latency;    /* finish time to measure latency */
269306 	unsigned long r_request_started; /* start time for mds request only,
270307 					    used to measure lease durations */
271308 
....@@ -282,8 +319,8 @@
282319 	int               r_num_fwd;    /* number of forward attempts */
283320 	int               r_resend_mds; /* mds to resend to next, if any*/
284321 	u32               r_sent_on_mseq; /* cap mseq request was sent at*/
322
+	u64		  r_deleg_ino;
285323 
286
-	struct kref       r_kref;
287324 	struct list_head  r_wait;
288325 	struct completion r_completion;
289326 	struct completion r_safe_completion;
....@@ -294,10 +331,8 @@
294331 	long long	  r_dir_release_cnt;
295332 	long long	  r_dir_ordered_cnt;
296333 	int		  r_readdir_cache_idx;
297
-	u32               r_readdir_offset;
298334 
299335 	struct ceph_cap_reservation r_caps_reservation;
300
-	int r_num_caps;
301336 };
302337 
303338 struct ceph_pool_perm {
....@@ -306,6 +341,35 @@
306341 	s64 pool;
307342 	size_t pool_ns_len;
308343 	char pool_ns[];
344
+};
345
+
346
+struct ceph_snapid_map {
347
+	struct rb_node node;
348
+	struct list_head lru;
349
+	atomic_t ref;
350
+	u64 snap;
351
+	dev_t dev;
352
+	unsigned long last_used;
353
+};
354
+
355
+/*
356
+ * node for list of quotarealm inodes that are not visible from the filesystem
357
+ * mountpoint, but required to handle, e.g. quotas.
358
+ */
359
+struct ceph_quotarealm_inode {
360
+	struct rb_node node;
361
+	u64 ino;
362
+	unsigned long timeout; /* last time a lookup failed for this inode */
363
+	struct mutex mutex;
364
+	struct inode *inode;
365
+};
366
+
367
+struct cap_wait {
368
+	struct list_head	list;
369
+	u64			ino;
370
+	pid_t			tgid;
371
+	int			need;
372
+	int			want;
309373 };
310374 
311375 /*
....@@ -323,10 +387,16 @@
323387 
324388 	struct ceph_mds_session **sessions;    /* NULL for mds if no session */
325389 	atomic_t		num_sessions;
326
-	int                     max_sessions;  /* len of s_mds_sessions */
390
+	int                     max_sessions;  /* len of sessions array */
327391 	int                     stopping;      /* true if shutting down */
328392 
329393 	atomic64_t		quotarealms_count; /* # realms with quota */
394
+	/*
395
+	 * We keep a list of inodes we don't see in the mountpoint but that we
396
+	 * need to track quota realms.
397
+	 */
398
+	struct rb_root		quotarealms_inodes;
399
+	struct mutex		quotarealms_inodes_mutex;
330400 
331401 	/*
332402 	 * snap_rwsem will cover cap linkage into snaprealms, and
....@@ -339,6 +409,7 @@
339409 	struct rw_semaphore     snap_rwsem;
340410 	struct rb_root          snap_realms;
341411 	struct list_head        snap_empty;
412
+	int			num_snap_realms;
342413 	spinlock_t              snap_empty_lock;  /* protect snap_empty */
343414 
344415 	u64                    last_tid;      /* most recent mds request */
....@@ -354,11 +425,13 @@
354425 
355426 	u64               last_cap_flush_tid;
356427 	struct list_head  cap_flush_list;
357
-	struct list_head  cap_dirty;        /* inodes with dirty caps */
358428 	struct list_head  cap_dirty_migrating; /* ...that are migration... */
359429 	int               num_cap_flushing; /* # caps we are flushing */
360430 	spinlock_t        cap_dirty_lock;   /* protects above items */
361431 	wait_queue_head_t cap_flushing_wq;
432
+
433
+	struct work_struct cap_reclaim_work;
434
+	atomic_t	   cap_reclaim_pending;
362435 
363436 	/*
364437 	 * Cap reservations
....@@ -374,15 +447,23 @@
374447 	spinlock_t	caps_list_lock;
375448 	struct		list_head caps_list; /* unused (reserved or
376449 						unreserved) */
450
+	struct		list_head cap_wait_list;
377451 	int		caps_total_count;    /* total caps allocated */
378452 	int		caps_use_count;      /* in use */
453
+	int		caps_use_max;	     /* max used caps */
379454 	int		caps_reserve_count;  /* unused, reserved */
380455 	int		caps_avail_count;    /* unused, unreserved */
381456 	int		caps_min_count;      /* keep at least this many
382457 						(unreserved) */
383
-	spinlock_t	  dentry_lru_lock;
384
-	struct list_head  dentry_lru;
385
-	int		  num_dentry;
458
+	spinlock_t	  dentry_list_lock;
459
+	struct list_head  dentry_leases;     /* fifo list */
460
+	struct list_head  dentry_dir_leases; /* lru list */
461
+
462
+	struct ceph_client_metric metric;
463
+
464
+	spinlock_t		snapid_map_lock;
465
+	struct rb_root		snapid_map_tree;
466
+	struct list_head	snapid_map_lru;
386467 
387468 	struct rw_semaphore     pool_perm_rwsem;
388469 	struct rb_root		pool_perm_tree;
....@@ -392,18 +473,16 @@
392473 
393474 extern const char *ceph_mds_op_name(int op);
394475 
476
+extern bool check_session_state(struct ceph_mds_session *s);
477
+void inc_session_sequence(struct ceph_mds_session *s);
478
+
395479 extern struct ceph_mds_session *
396480 __ceph_lookup_mds_session(struct ceph_mds_client *, int mds);
397481 
398
-static inline struct ceph_mds_session *
399
-ceph_get_mds_session(struct ceph_mds_session *s)
400
-{
401
-	refcount_inc(&s->s_ref);
402
-	return s;
403
-}
404
-
405482 extern const char *ceph_session_state_name(int s);
406483 
484
+extern struct ceph_mds_session *
485
+ceph_get_mds_session(struct ceph_mds_session *s);
407486 extern void ceph_put_mds_session(struct ceph_mds_session *s);
408487 
409488 extern int ceph_send_msg_mds(struct ceph_mds_client *mdsc,
....@@ -421,11 +500,14 @@
421500 					   struct inode *dir);
422501 extern struct ceph_mds_request *
423502 ceph_mdsc_create_request(struct ceph_mds_client *mdsc, int op, int mode);
424
-extern void ceph_mdsc_submit_request(struct ceph_mds_client *mdsc,
425
-				     struct ceph_mds_request *req);
503
+extern int ceph_mdsc_submit_request(struct ceph_mds_client *mdsc,
504
+				    struct inode *dir,
505
+				    struct ceph_mds_request *req);
426506 extern int ceph_mdsc_do_request(struct ceph_mds_client *mdsc,
427507 				struct inode *dir,
428508 				struct ceph_mds_request *req);
509
+extern void ceph_mdsc_release_dir_caps(struct ceph_mds_request *req);
510
+extern void ceph_mdsc_release_dir_caps_no_check(struct ceph_mds_request *req);
429511 static inline void ceph_mdsc_get_request(struct ceph_mds_request *req)
430512 {
431513 	kref_get(&req->r_kref);
....@@ -436,17 +518,34 @@
436518 	kref_put(&req->r_kref, ceph_mdsc_release_request);
437519 }
438520 
439
-extern void ceph_send_cap_releases(struct ceph_mds_client *mdsc,
440
-				   struct ceph_mds_session *session);
441
-
521
+extern void send_flush_mdlog(struct ceph_mds_session *s);
522
+extern void ceph_mdsc_iterate_sessions(struct ceph_mds_client *mdsc,
523
+				       void (*cb)(struct ceph_mds_session *),
524
+				       bool check_state);
525
+extern struct ceph_msg *ceph_create_session_msg(u32 op, u64 seq);
526
+extern void __ceph_queue_cap_release(struct ceph_mds_session *session,
527
+				    struct ceph_cap *cap);
528
+extern void ceph_flush_cap_releases(struct ceph_mds_client *mdsc,
529
+				    struct ceph_mds_session *session);
530
+extern void ceph_queue_cap_reclaim_work(struct ceph_mds_client *mdsc);
531
+extern void ceph_reclaim_caps_nr(struct ceph_mds_client *mdsc, int nr);
532
+extern int ceph_iterate_session_caps(struct ceph_mds_session *session,
533
+				     int (*cb)(struct inode *,
534
+					       struct ceph_cap *, void *),
535
+				     void *arg);
442536 extern void ceph_mdsc_pre_umount(struct ceph_mds_client *mdsc);
537
+
538
+static inline void ceph_mdsc_free_path(char *path, int len)
539
+{
540
+	if (!IS_ERR_OR_NULL(path))
541
+		__putname(path - (PATH_MAX - 1 - len));
542
+}
443543 
444544 extern char *ceph_mdsc_build_path(struct dentry *dentry, int *plen, u64 *base,
445545 				  int stop_on_nosnap);
446546 
447547 extern void __ceph_mdsc_drop_dentry_lease(struct dentry *dentry);
448548 extern void ceph_mdsc_lease_send_msg(struct ceph_mds_session *session,
449
-				     struct inode *inode,
450549 				     struct dentry *dentry, char action,
451550 				     u32 seq);
452551 
....@@ -463,4 +562,15 @@
463562 extern int ceph_trim_caps(struct ceph_mds_client *mdsc,
464563 			  struct ceph_mds_session *session,
465564 			  int max_caps);
565
+
566
+static inline int ceph_wait_on_async_create(struct inode *inode)
567
+{
568
+	struct ceph_inode_info *ci = ceph_inode(inode);
569
+
570
+	return wait_on_bit(&ci->i_ceph_flags, CEPH_ASYNC_CREATE_BIT,
571
+			   TASK_INTERRUPTIBLE);
572
+}
573
+
574
+extern u64 ceph_get_deleg_ino(struct ceph_mds_session *session);
575
+extern int ceph_restore_deleg_ino(struct ceph_mds_session *session, u64 ino);
466576 #endif