1 /****************************************************************************** 2 ******************************************************************************* 3 ** 4 ** Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved. 5 ** Copyright (C) 2004-2005 Red Hat, Inc. All rights reserved. 6 ** 7 ** This copyrighted material is made available to anyone wishing to use, 8 ** modify, copy, or redistribute it subject to the terms and conditions 9 ** of the GNU General Public License v.2. 10 ** 11 ******************************************************************************* 12 ******************************************************************************/ 13 14 #ifndef __DLM_INTERNAL_DOT_H__ 15 #define __DLM_INTERNAL_DOT_H__ 16 17 /* 18 * This is the main header file to be included in each DLM source file. 19 */ 20 21 #include <linux/module.h> 22 #include <linux/slab.h> 23 #include <linux/sched.h> 24 #include <linux/types.h> 25 #include <linux/ctype.h> 26 #include <linux/spinlock.h> 27 #include <linux/vmalloc.h> 28 #include <linux/list.h> 29 #include <linux/errno.h> 30 #include <linux/random.h> 31 #include <linux/delay.h> 32 #include <linux/socket.h> 33 #include <linux/kthread.h> 34 #include <linux/kobject.h> 35 #include <linux/kref.h> 36 #include <linux/kernel.h> 37 #include <linux/jhash.h> 38 #include <linux/miscdevice.h> 39 #include <linux/mutex.h> 40 #include <asm/semaphore.h> 41 #include <asm/uaccess.h> 42 43 #include <linux/dlm.h> 44 #include "config.h" 45 46 #define DLM_LOCKSPACE_LEN 64 47 48 /* Size of the temp buffer midcomms allocates on the stack. 49 We try to make this large enough so most messages fit. 50 FIXME: should sctp make this unnecessary? */ 51 52 #define DLM_INBUF_LEN 148 53 54 struct dlm_ls; 55 struct dlm_lkb; 56 struct dlm_rsb; 57 struct dlm_member; 58 struct dlm_lkbtable; 59 struct dlm_rsbtable; 60 struct dlm_dirtable; 61 struct dlm_direntry; 62 struct dlm_recover; 63 struct dlm_header; 64 struct dlm_message; 65 struct dlm_rcom; 66 struct dlm_mhandle; 67 68 #define log_print(fmt, args...) \ 69 printk(KERN_ERR "dlm: "fmt"\n" , ##args) 70 #define log_error(ls, fmt, args...) \ 71 printk(KERN_ERR "dlm: %s: " fmt "\n", (ls)->ls_name , ##args) 72 73 #define log_debug(ls, fmt, args...) \ 74 do { \ 75 if (dlm_config.ci_log_debug) \ 76 printk(KERN_DEBUG "dlm: %s: " fmt "\n", \ 77 (ls)->ls_name , ##args); \ 78 } while (0) 79 80 #define DLM_ASSERT(x, do) \ 81 { \ 82 if (!(x)) \ 83 { \ 84 printk(KERN_ERR "\nDLM: Assertion failed on line %d of file %s\n" \ 85 "DLM: assertion: \"%s\"\n" \ 86 "DLM: time = %lu\n", \ 87 __LINE__, __FILE__, #x, jiffies); \ 88 {do} \ 89 printk("\n"); \ 90 BUG(); \ 91 panic("DLM: Record message above and reboot.\n"); \ 92 } \ 93 } 94 95 #define DLM_FAKE_USER_AST ERR_PTR(-EINVAL) 96 97 98 struct dlm_direntry { 99 struct list_head list; 100 uint32_t master_nodeid; 101 uint16_t length; 102 char name[1]; 103 }; 104 105 struct dlm_dirtable { 106 struct list_head list; 107 rwlock_t lock; 108 }; 109 110 struct dlm_rsbtable { 111 struct list_head list; 112 struct list_head toss; 113 rwlock_t lock; 114 }; 115 116 struct dlm_lkbtable { 117 struct list_head list; 118 rwlock_t lock; 119 uint16_t counter; 120 }; 121 122 /* 123 * Lockspace member (per node in a ls) 124 */ 125 126 struct dlm_member { 127 struct list_head list; 128 int nodeid; 129 int weight; 130 }; 131 132 /* 133 * Save and manage recovery state for a lockspace. 134 */ 135 136 struct dlm_recover { 137 struct list_head list; 138 int *nodeids; 139 int node_count; 140 uint64_t seq; 141 }; 142 143 /* 144 * Pass input args to second stage locking function. 145 */ 146 147 struct dlm_args { 148 uint32_t flags; 149 void *astaddr; 150 long astparam; 151 void *bastaddr; 152 int mode; 153 struct dlm_lksb *lksb; 154 }; 155 156 157 /* 158 * Lock block 159 * 160 * A lock can be one of three types: 161 * 162 * local copy lock is mastered locally 163 * (lkb_nodeid is zero and DLM_LKF_MSTCPY is not set) 164 * process copy lock is mastered on a remote node 165 * (lkb_nodeid is non-zero and DLM_LKF_MSTCPY is not set) 166 * master copy master node's copy of a lock owned by remote node 167 * (lkb_nodeid is non-zero and DLM_LKF_MSTCPY is set) 168 * 169 * lkb_exflags: a copy of the most recent flags arg provided to dlm_lock or 170 * dlm_unlock. The dlm does not modify these or use any private flags in 171 * this field; it only contains DLM_LKF_ flags from dlm.h. These flags 172 * are sent as-is to the remote master when the lock is remote. 173 * 174 * lkb_flags: internal dlm flags (DLM_IFL_ prefix) from dlm_internal.h. 175 * Some internal flags are shared between the master and process nodes; 176 * these shared flags are kept in the lower two bytes. One of these 177 * flags set on the master copy will be propagated to the process copy 178 * and v.v. Other internal flags are private to the master or process 179 * node (e.g. DLM_IFL_MSTCPY). These are kept in the high two bytes. 180 * 181 * lkb_sbflags: status block flags. These flags are copied directly into 182 * the caller's lksb.sb_flags prior to the dlm_lock/dlm_unlock completion 183 * ast. All defined in dlm.h with DLM_SBF_ prefix. 184 * 185 * lkb_status: the lock status indicates which rsb queue the lock is 186 * on, grant, convert, or wait. DLM_LKSTS_ WAITING/GRANTED/CONVERT 187 * 188 * lkb_wait_type: the dlm message type (DLM_MSG_ prefix) for which a 189 * reply is needed. Only set when the lkb is on the lockspace waiters 190 * list awaiting a reply from a remote node. 191 * 192 * lkb_nodeid: when the lkb is a local copy, nodeid is 0; when the lkb 193 * is a master copy, nodeid specifies the remote lock holder, when the 194 * lkb is a process copy, the nodeid specifies the lock master. 195 */ 196 197 /* lkb_ast_type */ 198 199 #define AST_COMP 1 200 #define AST_BAST 2 201 202 /* lkb_status */ 203 204 #define DLM_LKSTS_WAITING 1 205 #define DLM_LKSTS_GRANTED 2 206 #define DLM_LKSTS_CONVERT 3 207 208 /* lkb_flags */ 209 210 #define DLM_IFL_MSTCPY 0x00010000 211 #define DLM_IFL_RESEND 0x00020000 212 #define DLM_IFL_DEAD 0x00040000 213 #define DLM_IFL_USER 0x00000001 214 #define DLM_IFL_ORPHAN 0x00000002 215 216 struct dlm_lkb { 217 struct dlm_rsb *lkb_resource; /* the rsb */ 218 struct kref lkb_ref; 219 int lkb_nodeid; /* copied from rsb */ 220 int lkb_ownpid; /* pid of lock owner */ 221 uint32_t lkb_id; /* our lock ID */ 222 uint32_t lkb_remid; /* lock ID on remote partner */ 223 uint32_t lkb_exflags; /* external flags from caller */ 224 uint32_t lkb_sbflags; /* lksb flags */ 225 uint32_t lkb_flags; /* internal flags */ 226 uint32_t lkb_lvbseq; /* lvb sequence number */ 227 228 int8_t lkb_status; /* granted, waiting, convert */ 229 int8_t lkb_rqmode; /* requested lock mode */ 230 int8_t lkb_grmode; /* granted lock mode */ 231 int8_t lkb_bastmode; /* requested mode */ 232 int8_t lkb_highbast; /* highest mode bast sent for */ 233 234 int8_t lkb_wait_type; /* type of reply waiting for */ 235 int8_t lkb_ast_type; /* type of ast queued for */ 236 237 struct list_head lkb_idtbl_list; /* lockspace lkbtbl */ 238 struct list_head lkb_statequeue; /* rsb g/c/w list */ 239 struct list_head lkb_rsb_lookup; /* waiting for rsb lookup */ 240 struct list_head lkb_wait_reply; /* waiting for remote reply */ 241 struct list_head lkb_astqueue; /* need ast to be sent */ 242 struct list_head lkb_ownqueue; /* list of locks for a process */ 243 244 char *lkb_lvbptr; 245 struct dlm_lksb *lkb_lksb; /* caller's status block */ 246 void *lkb_astaddr; /* caller's ast function */ 247 void *lkb_bastaddr; /* caller's bast function */ 248 long lkb_astparam; /* caller's ast arg */ 249 }; 250 251 252 struct dlm_rsb { 253 struct dlm_ls *res_ls; /* the lockspace */ 254 struct kref res_ref; 255 struct mutex res_mutex; 256 unsigned long res_flags; 257 int res_length; /* length of rsb name */ 258 int res_nodeid; 259 uint32_t res_lvbseq; 260 uint32_t res_hash; 261 uint32_t res_bucket; /* rsbtbl */ 262 unsigned long res_toss_time; 263 uint32_t res_first_lkid; 264 struct list_head res_lookup; /* lkbs waiting on first */ 265 struct list_head res_hashchain; /* rsbtbl */ 266 struct list_head res_grantqueue; 267 struct list_head res_convertqueue; 268 struct list_head res_waitqueue; 269 270 struct list_head res_root_list; /* used for recovery */ 271 struct list_head res_recover_list; /* used for recovery */ 272 int res_recover_locks_count; 273 274 char *res_lvbptr; 275 char res_name[1]; 276 }; 277 278 /* find_rsb() flags */ 279 280 #define R_MASTER 1 /* only return rsb if it's a master */ 281 #define R_CREATE 2 /* create/add rsb if not found */ 282 283 /* rsb_flags */ 284 285 enum rsb_flags { 286 RSB_MASTER_UNCERTAIN, 287 RSB_VALNOTVALID, 288 RSB_VALNOTVALID_PREV, 289 RSB_NEW_MASTER, 290 RSB_NEW_MASTER2, 291 RSB_RECOVER_CONVERT, 292 RSB_LOCKS_PURGED, 293 }; 294 295 static inline void rsb_set_flag(struct dlm_rsb *r, enum rsb_flags flag) 296 { 297 __set_bit(flag, &r->res_flags); 298 } 299 300 static inline void rsb_clear_flag(struct dlm_rsb *r, enum rsb_flags flag) 301 { 302 __clear_bit(flag, &r->res_flags); 303 } 304 305 static inline int rsb_flag(struct dlm_rsb *r, enum rsb_flags flag) 306 { 307 return test_bit(flag, &r->res_flags); 308 } 309 310 311 /* dlm_header is first element of all structs sent between nodes */ 312 313 #define DLM_HEADER_MAJOR 0x00030000 314 #define DLM_HEADER_MINOR 0x00000000 315 316 #define DLM_MSG 1 317 #define DLM_RCOM 2 318 319 struct dlm_header { 320 uint32_t h_version; 321 uint32_t h_lockspace; 322 uint32_t h_nodeid; /* nodeid of sender */ 323 uint16_t h_length; 324 uint8_t h_cmd; /* DLM_MSG, DLM_RCOM */ 325 uint8_t h_pad; 326 }; 327 328 329 #define DLM_MSG_REQUEST 1 330 #define DLM_MSG_CONVERT 2 331 #define DLM_MSG_UNLOCK 3 332 #define DLM_MSG_CANCEL 4 333 #define DLM_MSG_REQUEST_REPLY 5 334 #define DLM_MSG_CONVERT_REPLY 6 335 #define DLM_MSG_UNLOCK_REPLY 7 336 #define DLM_MSG_CANCEL_REPLY 8 337 #define DLM_MSG_GRANT 9 338 #define DLM_MSG_BAST 10 339 #define DLM_MSG_LOOKUP 11 340 #define DLM_MSG_REMOVE 12 341 #define DLM_MSG_LOOKUP_REPLY 13 342 343 struct dlm_message { 344 struct dlm_header m_header; 345 uint32_t m_type; /* DLM_MSG_ */ 346 uint32_t m_nodeid; 347 uint32_t m_pid; 348 uint32_t m_lkid; /* lkid on sender */ 349 uint32_t m_remid; /* lkid on receiver */ 350 uint32_t m_parent_lkid; 351 uint32_t m_parent_remid; 352 uint32_t m_exflags; 353 uint32_t m_sbflags; 354 uint32_t m_flags; 355 uint32_t m_lvbseq; 356 uint32_t m_hash; 357 int m_status; 358 int m_grmode; 359 int m_rqmode; 360 int m_bastmode; 361 int m_asts; 362 int m_result; /* 0 or -EXXX */ 363 char m_extra[0]; /* name or lvb */ 364 }; 365 366 367 #define DLM_RS_NODES 0x00000001 368 #define DLM_RS_NODES_ALL 0x00000002 369 #define DLM_RS_DIR 0x00000004 370 #define DLM_RS_DIR_ALL 0x00000008 371 #define DLM_RS_LOCKS 0x00000010 372 #define DLM_RS_LOCKS_ALL 0x00000020 373 #define DLM_RS_DONE 0x00000040 374 #define DLM_RS_DONE_ALL 0x00000080 375 376 #define DLM_RCOM_STATUS 1 377 #define DLM_RCOM_NAMES 2 378 #define DLM_RCOM_LOOKUP 3 379 #define DLM_RCOM_LOCK 4 380 #define DLM_RCOM_STATUS_REPLY 5 381 #define DLM_RCOM_NAMES_REPLY 6 382 #define DLM_RCOM_LOOKUP_REPLY 7 383 #define DLM_RCOM_LOCK_REPLY 8 384 385 struct dlm_rcom { 386 struct dlm_header rc_header; 387 uint32_t rc_type; /* DLM_RCOM_ */ 388 int rc_result; /* multi-purpose */ 389 uint64_t rc_id; /* match reply with request */ 390 uint64_t rc_seq; /* sender's ls_recover_seq */ 391 uint64_t rc_seq_reply; /* remote ls_recover_seq */ 392 char rc_buf[0]; 393 }; 394 395 struct rcom_config { 396 uint32_t rf_lvblen; 397 uint32_t rf_lsflags; 398 uint64_t rf_unused; 399 }; 400 401 struct rcom_lock { 402 uint32_t rl_ownpid; 403 uint32_t rl_lkid; 404 uint32_t rl_remid; 405 uint32_t rl_parent_lkid; 406 uint32_t rl_parent_remid; 407 uint32_t rl_exflags; 408 uint32_t rl_flags; 409 uint32_t rl_lvbseq; 410 int rl_result; 411 int8_t rl_rqmode; 412 int8_t rl_grmode; 413 int8_t rl_status; 414 int8_t rl_asts; 415 uint16_t rl_wait_type; 416 uint16_t rl_namelen; 417 char rl_name[DLM_RESNAME_MAXLEN]; 418 char rl_lvb[0]; 419 }; 420 421 struct dlm_ls { 422 struct list_head ls_list; /* list of lockspaces */ 423 dlm_lockspace_t *ls_local_handle; 424 uint32_t ls_global_id; /* global unique lockspace ID */ 425 uint32_t ls_exflags; 426 int ls_lvblen; 427 int ls_count; /* reference count */ 428 unsigned long ls_flags; /* LSFL_ */ 429 struct kobject ls_kobj; 430 431 struct dlm_rsbtable *ls_rsbtbl; 432 uint32_t ls_rsbtbl_size; 433 434 struct dlm_lkbtable *ls_lkbtbl; 435 uint32_t ls_lkbtbl_size; 436 437 struct dlm_dirtable *ls_dirtbl; 438 uint32_t ls_dirtbl_size; 439 440 struct mutex ls_waiters_mutex; 441 struct list_head ls_waiters; /* lkbs needing a reply */ 442 443 struct list_head ls_nodes; /* current nodes in ls */ 444 struct list_head ls_nodes_gone; /* dead node list, recovery */ 445 int ls_num_nodes; /* number of nodes in ls */ 446 int ls_low_nodeid; 447 int ls_total_weight; 448 int *ls_node_array; 449 450 struct dlm_rsb ls_stub_rsb; /* for returning errors */ 451 struct dlm_lkb ls_stub_lkb; /* for returning errors */ 452 struct dlm_message ls_stub_ms; /* for faking a reply */ 453 454 struct dentry *ls_debug_rsb_dentry; /* debugfs */ 455 struct dentry *ls_debug_waiters_dentry; /* debugfs */ 456 457 wait_queue_head_t ls_uevent_wait; /* user part of join/leave */ 458 int ls_uevent_result; 459 460 struct miscdevice ls_device; 461 462 /* recovery related */ 463 464 struct timer_list ls_timer; 465 struct task_struct *ls_recoverd_task; 466 struct mutex ls_recoverd_active; 467 spinlock_t ls_recover_lock; 468 uint32_t ls_recover_status; /* DLM_RS_ */ 469 uint64_t ls_recover_seq; 470 struct dlm_recover *ls_recover_args; 471 struct rw_semaphore ls_in_recovery; /* block local requests */ 472 struct list_head ls_requestqueue;/* queue remote requests */ 473 struct mutex ls_requestqueue_mutex; 474 char *ls_recover_buf; 475 int ls_recover_nodeid; /* for debugging */ 476 uint64_t ls_rcom_seq; 477 spinlock_t ls_rcom_spin; 478 struct list_head ls_recover_list; 479 spinlock_t ls_recover_list_lock; 480 int ls_recover_list_count; 481 wait_queue_head_t ls_wait_general; 482 struct mutex ls_clear_proc_locks; 483 484 struct list_head ls_root_list; /* root resources */ 485 struct rw_semaphore ls_root_sem; /* protect root_list */ 486 487 int ls_namelen; 488 char ls_name[1]; 489 }; 490 491 #define LSFL_WORK 0 492 #define LSFL_RUNNING 1 493 #define LSFL_RECOVERY_STOP 2 494 #define LSFL_RCOM_READY 3 495 #define LSFL_RCOM_WAIT 4 496 #define LSFL_UEVENT_WAIT 5 497 498 /* much of this is just saving user space pointers associated with the 499 lock that we pass back to the user lib with an ast */ 500 501 struct dlm_user_args { 502 struct dlm_user_proc *proc; /* each process that opens the lockspace 503 device has private data 504 (dlm_user_proc) on the struct file, 505 the process's locks point back to it*/ 506 struct dlm_lksb lksb; 507 int old_mode; 508 int update_user_lvb; 509 struct dlm_lksb __user *user_lksb; 510 void __user *castparam; 511 void __user *castaddr; 512 void __user *bastparam; 513 void __user *bastaddr; 514 }; 515 516 #define DLM_PROC_FLAGS_CLOSING 1 517 #define DLM_PROC_FLAGS_COMPAT 2 518 519 /* locks list is kept so we can remove all a process's locks when it 520 exits (or orphan those that are persistent) */ 521 522 struct dlm_user_proc { 523 dlm_lockspace_t *lockspace; 524 unsigned long flags; /* DLM_PROC_FLAGS */ 525 struct list_head asts; 526 spinlock_t asts_spin; 527 struct list_head locks; 528 spinlock_t locks_spin; 529 struct list_head unlocking; 530 wait_queue_head_t wait; 531 }; 532 533 static inline int dlm_locking_stopped(struct dlm_ls *ls) 534 { 535 return !test_bit(LSFL_RUNNING, &ls->ls_flags); 536 } 537 538 static inline int dlm_recovery_stopped(struct dlm_ls *ls) 539 { 540 return test_bit(LSFL_RECOVERY_STOP, &ls->ls_flags); 541 } 542 543 static inline int dlm_no_directory(struct dlm_ls *ls) 544 { 545 return (ls->ls_exflags & DLM_LSFL_NODIR) ? 1 : 0; 546 } 547 548 #endif /* __DLM_INTERNAL_DOT_H__ */ 549 550