fs/orangefs/orangefs-kernel.h

   1 /*
   2  * (C) 2001 Clemson University and The University of Chicago
   3  *
   4  * See COPYING in top-level directory.
   5  */
   6
   7 /*
   8  *  The ORANGEFS Linux kernel support allows ORANGEFS volumes to be mounted and
   9  *  accessed through the Linux VFS (i.e. using standard I/O system calls).
  10  *  This support is only needed on clients that wish to mount the file system.
  11  *
  12  */
  13
  14 /*
  15  *  Declarations and macros for the ORANGEFS Linux kernel support.
  16  */
  17
  18 #ifndef __ORANGEFSKERNEL_H
  19 #define __ORANGEFSKERNEL_H
  20
  21 #include <linux/kernel.h>
  22 #include <linux/moduleparam.h>
  23 #include <linux/statfs.h>
  24 #include <linux/backing-dev.h>
  25 #include <linux/device.h>
  26 #include <linux/mpage.h>
  27 #include <linux/namei.h>
  28 #include <linux/errno.h>
  29 #include <linux/init.h>
  30 #include <linux/module.h>
  31 #include <linux/slab.h>
  32 #include <linux/types.h>
  33 #include <linux/fs.h>
  34 #include <linux/vmalloc.h>
  35
  36 #include <linux/aio.h>
  37 #include <linux/posix_acl.h>
  38 #include <linux/posix_acl_xattr.h>
  39 #include <linux/compat.h>
  40 #include <linux/mount.h>
  41 #include <linux/uaccess.h>
  42 #include <linux/atomic.h>
  43 #include <linux/uio.h>
  44 #include <linux/sched.h>
  45 #include <linux/mm.h>
  46 #include <linux/wait.h>
  47 #include <linux/dcache.h>
  48 #include <linux/pagemap.h>
  49 #include <linux/poll.h>
  50 #include <linux/rwsem.h>
  51 #include <linux/xattr.h>
  52 #include <linux/exportfs.h>
  53
  54 #include <asm/unaligned.h>
  55
  56 #include "orangefs-dev-proto.h"
  57
  58 #ifdef ORANGEFS_KERNEL_DEBUG
  59 #define ORANGEFS_DEFAULT_OP_TIMEOUT_SECS       10
  60 #else
  61 #define ORANGEFS_DEFAULT_OP_TIMEOUT_SECS       20
  62 #endif
  63
  64 #define ORANGEFS_BUFMAP_WAIT_TIMEOUT_SECS   30
  65
  66 #define ORANGEFS_DEFAULT_SLOT_TIMEOUT_SECS     900      /* 15 minutes */
  67
  68 #define ORANGEFS_REQDEVICE_NAME          "pvfs2-req"
  69
  70 #define ORANGEFS_DEVREQ_MAGIC             0x20030529
  71 #define ORANGEFS_LINK_MAX                 0x000000FF
  72 #define ORANGEFS_PURGE_RETRY_COUNT     0x00000005
  73 #define ORANGEFS_SEEK_END              0x00000002
  74 #define ORANGEFS_MAX_NUM_OPTIONS          0x00000004
  75 #define ORANGEFS_MAX_MOUNT_OPT_LEN        0x00000080
  76 #define ORANGEFS_MAX_FSKEY_LEN            64
  77
  78 #define MAX_DEV_REQ_UPSIZE (2 * sizeof(__s32) +   \
  79 sizeof(__u64) + sizeof(struct orangefs_upcall_s))
  80 #define MAX_DEV_REQ_DOWNSIZE (2 * sizeof(__s32) + \
  81 sizeof(__u64) + sizeof(struct orangefs_downcall_s))
  82
  83 /* borrowed from irda.h */
  84 #ifndef MSECS_TO_JIFFIES
  85 #define MSECS_TO_JIFFIES(ms) (((ms)*HZ+999)/1000)
  86 #endif
  87
  88 /*
  89  * valid orangefs kernel operation states
  90  *
  91  * unknown  - op was just initialized
  92  * waiting  - op is on request_list (upward bound)
  93  * inprogr  - op is in progress (waiting for downcall)
  94  * serviced - op has matching downcall; ok
  95  * purged   - op has to start a timer since client-core
  96  *            exited uncleanly before servicing op
  97  */
  98 enum orangefs_vfs_op_states {
  99         OP_VFS_STATE_UNKNOWN = 0,
 100         OP_VFS_STATE_WAITING = 1,
 101         OP_VFS_STATE_INPROGR = 2,
 102         OP_VFS_STATE_SERVICED = 4,
 103         OP_VFS_STATE_PURGED = 8,
 104 };
 105
 106 #define get_op(op)                                      \
 107         do {                                            \
 108                 atomic_inc(&(op)->ref_count);   \
 109                 gossip_debug(GOSSIP_DEV_DEBUG,  \
 110                         "(get) Alloced OP (%p:%llu)\n", \
 111                         op,                             \
 112                         llu((op)->tag));                \
 113         } while (0)
 114
 115 #define put_op(op)                                                      \
 116         do {                                                            \
 117                 if (atomic_sub_and_test(1, &(op)->ref_count) == 1) {  \
 118                         gossip_debug(GOSSIP_DEV_DEBUG,          \
 119                                 "(put) Releasing OP (%p:%llu)\n",       \
 120                                 op,                                     \
 121                                 llu((op)->tag));                        \
 122                         op_release(op);                                 \
 123                         }                                               \
 124         } while (0)
 125
 126 #define op_wait(op) (atomic_read(&(op)->ref_count) <= 2 ? 0 : 1)
 127
 128 /*
 129  * Defines for controlling whether I/O upcalls are for async or sync operations
 130  */
 131 enum ORANGEFS_async_io_type {
 132         ORANGEFS_VFS_SYNC_IO = 0,
 133         ORANGEFS_VFS_ASYNC_IO = 1,
 134 };
 135
 136 /*
 137  * An array of client_debug_mask will be built to hold debug keyword/mask
 138  * values fetched from userspace.
 139  */
 140 struct client_debug_mask {
 141         char *keyword;
 142         __u64 mask1;
 143         __u64 mask2;
 144 };
 145
 146 /*
 147  * orangefs kernel memory related flags
 148  */
 149
 150 #if ((defined ORANGEFS_KERNEL_DEBUG) && (defined CONFIG_DEBUG_SLAB))
 151 #define ORANGEFS_CACHE_CREATE_FLAGS SLAB_RED_ZONE
 152 #else
 153 #define ORANGEFS_CACHE_CREATE_FLAGS 0
 154 #endif /* ((defined ORANGEFS_KERNEL_DEBUG) && (defined CONFIG_DEBUG_SLAB)) */
 155
 156 #define ORANGEFS_CACHE_ALLOC_FLAGS (GFP_KERNEL)
 157 #define ORANGEFS_GFP_FLAGS (GFP_KERNEL)
 158 #define ORANGEFS_BUFMAP_GFP_FLAGS (GFP_KERNEL)
 159
 160 /* orangefs xattr and acl related defines */
 161 #define ORANGEFS_XATTR_INDEX_POSIX_ACL_ACCESS  1
 162 #define ORANGEFS_XATTR_INDEX_POSIX_ACL_DEFAULT 2
 163 #define ORANGEFS_XATTR_INDEX_TRUSTED           3
 164 #define ORANGEFS_XATTR_INDEX_DEFAULT           4
 165
 166 #define ORANGEFS_XATTR_NAME_ACL_ACCESS XATTR_NAME_POSIX_ACL_ACCESS
 167 #define ORANGEFS_XATTR_NAME_ACL_DEFAULT XATTR_NAME_POSIX_ACL_DEFAULT
 168 #define ORANGEFS_XATTR_NAME_TRUSTED_PREFIX "trusted."
 169 #define ORANGEFS_XATTR_NAME_DEFAULT_PREFIX ""
 170
 171 /* these functions are defined in orangefs-utils.c */
 172 int orangefs_prepare_cdm_array(char *debug_array_string);
 173 int orangefs_prepare_debugfs_help_string(int);
 174
 175 /* defined in orangefs-debugfs.c */
 176 int orangefs_client_debug_init(void);
 177
 178 void debug_string_to_mask(char *, void *, int);
 179 void do_c_mask(int, char *, struct client_debug_mask **);
 180 void do_k_mask(int, char *, __u64 **);
 181
 182 void debug_mask_to_string(void *, int);
 183 void do_k_string(void *, int);
 184 void do_c_string(void *, int);
 185 int check_amalgam_keyword(void *, int);
 186 int keyword_is_amalgam(char *);
 187
 188 /*these variables are defined in orangefs-mod.c */
 189 extern char kernel_debug_string[ORANGEFS_MAX_DEBUG_STRING_LEN];
 190 extern char client_debug_string[ORANGEFS_MAX_DEBUG_STRING_LEN];
 191 extern char client_debug_array_string[ORANGEFS_MAX_DEBUG_STRING_LEN];
 192 extern unsigned int kernel_mask_set_mod_init;
 193
 194 extern int orangefs_init_acl(struct inode *inode, struct inode *dir);
 195 extern const struct xattr_handler *orangefs_xattr_handlers[];
 196
 197 extern struct posix_acl *orangefs_get_acl(struct inode *inode, int type);
 198 extern int orangefs_set_acl(struct inode *inode, struct posix_acl *acl, int type);
 199
 200 /*
 201  * Redefine xtvec structure so that we could move helper functions out of
 202  * the define
 203  */
 204 struct xtvec {
 205         __kernel_off_t xtv_off;         /* must be off_t */
 206         __kernel_size_t xtv_len;        /* must be size_t */
 207 };
 208
 209 /*
 210  * orangefs data structures
 211  */
 212 struct orangefs_kernel_op_s {
 213         enum orangefs_vfs_op_states op_state;
 214         __u64 tag;
 215
 216         /*
 217          * Set uses_shared_memory to 1 if this operation uses shared memory.
 218          * If true, then a retry on the op must also get a new shared memory
 219          * buffer and re-populate it.
 220          */
 221         int uses_shared_memory;
 222
 223         struct orangefs_upcall_s upcall;
 224         struct orangefs_downcall_s downcall;
 225
 226         wait_queue_head_t waitq;
 227         spinlock_t lock;
 228
 229         int io_completed;
 230         wait_queue_head_t io_completion_waitq;
 231
 232         atomic_t ref_count;
 233
 234         /* VFS aio fields */
 235
 236         /* used by the async I/O code to stash the orangefs_kiocb_s structure */
 237         void *priv;
 238
 239         int attempts;
 240
 241         struct list_head list;
 242 };
 243
 244 #define set_op_state_waiting(op)     ((op)->op_state = OP_VFS_STATE_WAITING)
 245 #define set_op_state_inprogress(op)  ((op)->op_state = OP_VFS_STATE_INPROGR)
 246 static inline void set_op_state_serviced(struct orangefs_kernel_op_s *op)
 247 {
 248         op->op_state = OP_VFS_STATE_SERVICED;
 249         wake_up_interruptible(&op->waitq);
 250 }
 251 static inline void set_op_state_purged(struct orangefs_kernel_op_s *op)
 252 {
 253         op->op_state |= OP_VFS_STATE_PURGED;
 254         wake_up_interruptible(&op->waitq);
 255 }
 256
 257 #define op_state_waiting(op)     ((op)->op_state & OP_VFS_STATE_WAITING)
 258 #define op_state_in_progress(op) ((op)->op_state & OP_VFS_STATE_INPROGR)
 259 #define op_state_serviced(op)    ((op)->op_state & OP_VFS_STATE_SERVICED)
 260 #define op_state_purged(op)      ((op)->op_state & OP_VFS_STATE_PURGED)
 261
 262 /* per inode private orangefs info */
 263 struct orangefs_inode_s {
 264         struct orangefs_object_kref refn;
 265         char link_target[ORANGEFS_NAME_MAX];
 266         __s64 blksize;
 267         /*
 268          * Reading/Writing Extended attributes need to acquire the appropriate
 269          * reader/writer semaphore on the orangefs_inode_s structure.
 270          */
 271         struct rw_semaphore xattr_sem;
 272
 273         struct inode vfs_inode;
 274         sector_t last_failed_block_index_read;
 275
 276         /*
 277          * State of in-memory attributes not yet flushed to disk associated
 278          * with this object
 279          */
 280         unsigned long pinode_flags;
 281 };
 282
 283 #define P_ATIME_FLAG 0
 284 #define P_MTIME_FLAG 1
 285 #define P_CTIME_FLAG 2
 286 #define P_MODE_FLAG  3
 287
 288 #define ClearAtimeFlag(pinode) clear_bit(P_ATIME_FLAG, &(pinode)->pinode_flags)
 289 #define SetAtimeFlag(pinode)   set_bit(P_ATIME_FLAG, &(pinode)->pinode_flags)
 290 #define AtimeFlag(pinode)      test_bit(P_ATIME_FLAG, &(pinode)->pinode_flags)
 291
 292 #define ClearMtimeFlag(pinode) clear_bit(P_MTIME_FLAG, &(pinode)->pinode_flags)
 293 #define SetMtimeFlag(pinode)   set_bit(P_MTIME_FLAG, &(pinode)->pinode_flags)
 294 #define MtimeFlag(pinode)      test_bit(P_MTIME_FLAG, &(pinode)->pinode_flags)
 295
 296 #define ClearCtimeFlag(pinode) clear_bit(P_CTIME_FLAG, &(pinode)->pinode_flags)
 297 #define SetCtimeFlag(pinode)   set_bit(P_CTIME_FLAG, &(pinode)->pinode_flags)
 298 #define CtimeFlag(pinode)      test_bit(P_CTIME_FLAG, &(pinode)->pinode_flags)
 299
 300 #define ClearModeFlag(pinode) clear_bit(P_MODE_FLAG, &(pinode)->pinode_flags)
 301 #define SetModeFlag(pinode)   set_bit(P_MODE_FLAG, &(pinode)->pinode_flags)
 302 #define ModeFlag(pinode)      test_bit(P_MODE_FLAG, &(pinode)->pinode_flags)
 303
 304 /* per superblock private orangefs info */
 305 struct orangefs_sb_info_s {
 306         struct orangefs_khandle root_khandle;
 307         __s32 fs_id;
 308         int id;
 309         int flags;
 310 #define ORANGEFS_OPT_INTR       0x01
 311 #define ORANGEFS_OPT_LOCAL_LOCK 0x02
 312         char devname[ORANGEFS_MAX_SERVER_ADDR_LEN];
 313         struct super_block *sb;
 314         int mount_pending;
 315         struct list_head list;
 316 };
 317
 318 /*
 319  * structure that holds the state of any async I/O operation issued
 320  * through the VFS. Needed especially to handle cancellation requests
 321  * or even completion notification so that the VFS client-side daemon
 322  * can free up its vfs_request slots.
 323  */
 324 struct orangefs_kiocb_s {
 325         /* the pointer to the task that initiated the AIO */
 326         struct task_struct *tsk;
 327
 328         /* pointer to the kiocb that kicked this operation */
 329         struct kiocb *kiocb;
 330
 331         /* buffer index that was used for the I/O */
 332         struct orangefs_bufmap *bufmap;
 333         int buffer_index;
 334
 335         /* orangefs kernel operation type */
 336         struct orangefs_kernel_op_s *op;
 337
 338         /* The user space buffers from/to which I/O is being staged */
 339         struct iovec *iov;
 340
 341         /* number of elements in the iovector */
 342         unsigned long nr_segs;
 343
 344         /* set to indicate the type of the operation */
 345         int rw;
 346
 347         /* file offset */
 348         loff_t offset;
 349
 350         /* and the count in bytes */
 351         size_t bytes_to_be_copied;
 352
 353         ssize_t bytes_copied;
 354         int needs_cleanup;
 355 };
 356
 357 struct orangefs_stats {
 358         unsigned long cache_hits;
 359         unsigned long cache_misses;
 360         unsigned long reads;
 361         unsigned long writes;
 362 };
 363
 364 extern struct orangefs_stats g_orangefs_stats;
 365
 366 /*
 367  * NOTE: See Documentation/filesystems/porting for information
 368  * on implementing FOO_I and properly accessing fs private data
 369  */
 370 static inline struct orangefs_inode_s *ORANGEFS_I(struct inode *inode)
 371 {
 372         return container_of(inode, struct orangefs_inode_s, vfs_inode);
 373 }
 374
 375 static inline struct orangefs_sb_info_s *ORANGEFS_SB(struct super_block *sb)
 376 {
 377         return (struct orangefs_sb_info_s *) sb->s_fs_info;
 378 }
 379
 380 /* ino_t descends from "unsigned long", 8 bytes, 64 bits. */
 381 static inline ino_t orangefs_khandle_to_ino(struct orangefs_khandle *khandle)
 382 {
 383         union {
 384                 unsigned char u[8];
 385                 __u64 ino;
 386         } ihandle;
 387
 388         ihandle.u[0] = khandle->u[0] ^ khandle->u[4];
 389         ihandle.u[1] = khandle->u[1] ^ khandle->u[5];
 390         ihandle.u[2] = khandle->u[2] ^ khandle->u[6];
 391         ihandle.u[3] = khandle->u[3] ^ khandle->u[7];
 392         ihandle.u[4] = khandle->u[12] ^ khandle->u[8];
 393         ihandle.u[5] = khandle->u[13] ^ khandle->u[9];
 394         ihandle.u[6] = khandle->u[14] ^ khandle->u[10];
 395         ihandle.u[7] = khandle->u[15] ^ khandle->u[11];
 396
 397         return ihandle.ino;
 398 }
 399
 400 static inline struct orangefs_khandle *get_khandle_from_ino(struct inode *inode)
 401 {
 402         return &(ORANGEFS_I(inode)->refn.khandle);
 403 }
 404
 405 static inline __s32 get_fsid_from_ino(struct inode *inode)
 406 {
 407         return ORANGEFS_I(inode)->refn.fs_id;
 408 }
 409
 410 static inline ino_t get_ino_from_khandle(struct inode *inode)
 411 {
 412         struct orangefs_khandle *khandle;
 413         ino_t ino;
 414
 415         khandle = get_khandle_from_ino(inode);
 416         ino = orangefs_khandle_to_ino(khandle);
 417         return ino;
 418 }
 419
 420 static inline ino_t get_parent_ino_from_dentry(struct dentry *dentry)
 421 {
 422         return get_ino_from_khandle(dentry->d_parent->d_inode);
 423 }
 424
 425 static inline int is_root_handle(struct inode *inode)
 426 {
 427         gossip_debug(GOSSIP_DCACHE_DEBUG,
 428                      "%s: root handle: %pU, this handle: %pU:\n",
 429                      __func__,
 430                      &ORANGEFS_SB(inode->i_sb)->root_khandle,
 431                      get_khandle_from_ino(inode));
 432
 433         if (ORANGEFS_khandle_cmp(&(ORANGEFS_SB(inode->i_sb)->root_khandle),
 434                              get_khandle_from_ino(inode)))
 435                 return 0;
 436         else
 437                 return 1;
 438 }
 439
 440 static inline int match_handle(struct orangefs_khandle resp_handle,
 441                                struct inode *inode)
 442 {
 443         gossip_debug(GOSSIP_DCACHE_DEBUG,
 444                      "%s: one handle: %pU, another handle:%pU:\n",
 445                      __func__,
 446                      &resp_handle,
 447                      get_khandle_from_ino(inode));
 448
 449         if (ORANGEFS_khandle_cmp(&resp_handle, get_khandle_from_ino(inode)))
 450                 return 0;
 451         else
 452                 return 1;
 453 }
 454
 455 /*
 456  * defined in orangefs-cache.c
 457  */
 458 int op_cache_initialize(void);
 459 int op_cache_finalize(void);
 460 struct orangefs_kernel_op_s *op_alloc(__s32 type);
 461 char *get_opname_string(struct orangefs_kernel_op_s *new_op);
 462 void op_release(struct orangefs_kernel_op_s *op);
 463
 464 int dev_req_cache_initialize(void);
 465 int dev_req_cache_finalize(void);
 466 void *dev_req_alloc(void);
 467 void dev_req_release(void *);
 468
 469 int orangefs_inode_cache_initialize(void);
 470 int orangefs_inode_cache_finalize(void);
 471
 472 int kiocb_cache_initialize(void);
 473 int kiocb_cache_finalize(void);
 474 struct orangefs_kiocb_s *kiocb_alloc(void);
 475 void kiocb_release(struct orangefs_kiocb_s *ptr);
 476
 477 /*
 478  * defined in orangefs-mod.c
 479  */
 480 void purge_inprogress_ops(void);
 481
 482 /*
 483  * defined in waitqueue.c
 484  */
 485 void purge_waiting_ops(void);
 486
 487 /*
 488  * defined in super.c
 489  */
 490 struct dentry *orangefs_mount(struct file_system_type *fst,
 491                            int flags,
 492                            const char *devname,
 493                            void *data);
 494
 495 void orangefs_kill_sb(struct super_block *sb);
 496 int orangefs_remount(struct super_block *sb);
 497
 498 int fsid_key_table_initialize(void);
 499 void fsid_key_table_finalize(void);
 500
 501 /*
 502  * defined in inode.c
 503  */
 504 __u32 convert_to_orangefs_mask(unsigned long lite_mask);
 505 struct inode *orangefs_new_inode(struct super_block *sb,
 506                               struct inode *dir,
 507                               int mode,
 508                               dev_t dev,
 509                               struct orangefs_object_kref *ref);
 510
 511 int orangefs_setattr(struct dentry *dentry, struct iattr *iattr);
 512
 513 int orangefs_getattr(struct vfsmount *mnt,
 514                   struct dentry *dentry,
 515                   struct kstat *kstat);
 516
 517 /*
 518  * defined in xattr.c
 519  */
 520 int orangefs_setxattr(struct dentry *dentry,
 521                    const char *name,
 522                    const void *value,
 523                    size_t size,
 524                    int flags);
 525
 526 ssize_t orangefs_getxattr(struct dentry *dentry,
 527                        const char *name,
 528                        void *buffer,
 529                        size_t size);
 530
 531 ssize_t orangefs_listxattr(struct dentry *dentry, char *buffer, size_t size);
 532
 533 /*
 534  * defined in namei.c
 535  */
 536 struct inode *orangefs_iget(struct super_block *sb,
 537                          struct orangefs_object_kref *ref);
 538
 539 ssize_t orangefs_inode_read(struct inode *inode,
 540                             struct iov_iter *iter,
 541                             loff_t *offset,
 542                             loff_t readahead_size);
 543
 544 /*
 545  * defined in devorangefs-req.c
 546  */
 547 int orangefs_dev_init(void);
 548 void orangefs_dev_cleanup(void);
 549 int is_daemon_in_service(void);
 550 int fs_mount_pending(__s32 fsid);
 551
 552 /*
 553  * defined in orangefs-utils.c
 554  */
 555 __s32 fsid_of_op(struct orangefs_kernel_op_s *op);
 556
 557 int orangefs_flush_inode(struct inode *inode);
 558
 559 ssize_t orangefs_inode_getxattr(struct inode *inode,
 560                              const char *prefix,
 561                              const char *name,
 562                              void *buffer,
 563                              size_t size);
 564
 565 int orangefs_inode_setxattr(struct inode *inode,
 566                          const char *prefix,
 567                          const char *name,
 568                          const void *value,
 569                          size_t size,
 570                          int flags);
 571
 572 int orangefs_inode_getattr(struct inode *inode, __u32 mask);
 573
 574 int orangefs_inode_setattr(struct inode *inode, struct iattr *iattr);
 575
 576 void orangefs_op_initialize(struct orangefs_kernel_op_s *op);
 577
 578 void orangefs_make_bad_inode(struct inode *inode);
 579
 580 void orangefs_block_signals(sigset_t *);
 581
 582 void orangefs_set_signals(sigset_t *);
 583
 584 int orangefs_unmount_sb(struct super_block *sb);
 585
 586 int orangefs_cancel_op_in_progress(__u64 tag);
 587
 588 static inline __u64 orangefs_convert_time_field(const struct timespec *ts)
 589 {
 590         return (__u64)ts->tv_sec;
 591 }
 592
 593 int orangefs_normalize_to_errno(__s32 error_code);
 594
 595 extern struct mutex devreq_mutex;
 596 extern struct mutex request_mutex;
 597 extern int debug;
 598 extern int op_timeout_secs;
 599 extern int slot_timeout_secs;
 600 extern struct list_head orangefs_superblocks;
 601 extern spinlock_t orangefs_superblocks_lock;
 602 extern struct list_head orangefs_request_list;
 603 extern spinlock_t orangefs_request_list_lock;
 604 extern wait_queue_head_t orangefs_request_list_waitq;
 605 extern struct list_head *htable_ops_in_progress;
 606 extern spinlock_t htable_ops_in_progress_lock;
 607 extern int hash_table_size;
 608
 609 extern const struct address_space_operations orangefs_address_operations;
 610 extern struct backing_dev_info orangefs_backing_dev_info;
 611 extern struct inode_operations orangefs_file_inode_operations;
 612 extern const struct file_operations orangefs_file_operations;
 613 extern struct inode_operations orangefs_symlink_inode_operations;
 614 extern struct inode_operations orangefs_dir_inode_operations;
 615 extern const struct file_operations orangefs_dir_operations;
 616 extern const struct dentry_operations orangefs_dentry_operations;
 617 extern const struct file_operations orangefs_devreq_file_operations;
 618
 619 extern wait_queue_head_t orangefs_bufmap_init_waitq;
 620
 621 /*
 622  * misc convenience macros
 623  */
 624
 625 #define ORANGEFS_OP_INTERRUPTIBLE 1   /* service_operation() is interruptible */
 626 #define ORANGEFS_OP_PRIORITY      2   /* service_operation() is high priority */
 627 #define ORANGEFS_OP_CANCELLATION  4   /* this is a cancellation */
 628 #define ORANGEFS_OP_NO_SEMAPHORE  8   /* don't acquire semaphore */
 629 #define ORANGEFS_OP_ASYNC         16  /* Queue it, but don't wait */
 630
 631 int service_operation(struct orangefs_kernel_op_s *op,
 632                       const char *op_name,
 633                       int flags);
 634
 635 /*
 636  * handles two possible error cases, depending on context.
 637  *
 638  * by design, our vfs i/o errors need to be handled in one of two ways,
 639  * depending on where the error occured.
 640  *
 641  * if the error happens in the waitqueue code because we either timed
 642  * out or a signal was raised while waiting, we need to cancel the
 643  * userspace i/o operation and free the op manually.  this is done to
 644  * avoid having the device start writing application data to our shared
 645  * bufmap pages without us expecting it.
 646  *
 647  * FIXME: POSSIBLE OPTIMIZATION:
 648  * However, if we timed out or if we got a signal AND our upcall was never
 649  * picked off the queue (i.e. we were in OP_VFS_STATE_WAITING), then we don't
 650  * need to send a cancellation upcall. The way we can handle this is
 651  * set error_exit to 2 in such cases and 1 whenever cancellation has to be
 652  * sent and have handle_error
 653  * take care of this situation as well..
 654  *
 655  * if a orangefs sysint level error occured and i/o has been completed,
 656  * there is no need to cancel the operation, as the user has finished
 657  * using the bufmap page and so there is no danger in this case.  in
 658  * this case, we wake up the device normally so that it may free the
 659  * op, as normal.
 660  *
 661  * note the only reason this is a macro is because both read and write
 662  * cases need the exact same handling code.
 663  */
 664 #define handle_io_error()                                       \
 665 do {                                                            \
 666         if (!op_state_serviced(new_op)) {                       \
 667                 orangefs_cancel_op_in_progress(new_op->tag);    \
 668                 op_release(new_op);                             \
 669         } else {                                                \
 670                 wake_up_daemon_for_return(new_op);              \
 671         }                                                       \
 672         new_op = NULL;                                          \
 673         orangefs_bufmap_put(bufmap, buffer_index);                              \
 674         buffer_index = -1;                                      \
 675 } while (0)
 676
 677 #define get_interruptible_flag(inode) \
 678         ((ORANGEFS_SB(inode->i_sb)->flags & ORANGEFS_OPT_INTR) ? \
 679                 ORANGEFS_OP_INTERRUPTIBLE : 0)
 680
 681 #define add_orangefs_sb(sb)                                             \
 682 do {                                                                    \
 683         gossip_debug(GOSSIP_SUPER_DEBUG,                                \
 684                      "Adding SB %p to orangefs superblocks\n",          \
 685                      ORANGEFS_SB(sb));                                  \
 686         spin_lock(&orangefs_superblocks_lock);                          \
 687         list_add_tail(&ORANGEFS_SB(sb)->list, &orangefs_superblocks);           \
 688         spin_unlock(&orangefs_superblocks_lock); \
 689 } while (0)
 690
 691 #define remove_orangefs_sb(sb)                                          \
 692 do {                                                                    \
 693         struct list_head *tmp = NULL;                                   \
 694         struct list_head *tmp_safe = NULL;                              \
 695         struct orangefs_sb_info_s *orangefs_sb = NULL;                  \
 696                                                                         \
 697         spin_lock(&orangefs_superblocks_lock);                          \
 698         list_for_each_safe(tmp, tmp_safe, &orangefs_superblocks) {              \
 699                 orangefs_sb = list_entry(tmp,                           \
 700                                       struct orangefs_sb_info_s,                \
 701                                       list);                            \
 702                 if (orangefs_sb && (orangefs_sb->sb == sb)) {                   \
 703                         gossip_debug(GOSSIP_SUPER_DEBUG,                \
 704                             "Removing SB %p from orangefs superblocks\n",       \
 705                         orangefs_sb);                                   \
 706                         list_del(&orangefs_sb->list);                   \
 707                         break;                                          \
 708                 }                                                       \
 709         }                                                               \
 710         spin_unlock(&orangefs_superblocks_lock);                                \
 711 } while (0)
 712
 713 #define orangefs_lock_inode(inode) spin_lock(&inode->i_lock)
 714 #define orangefs_unlock_inode(inode) spin_unlock(&inode->i_lock)
 715
 716 #define fill_default_sys_attrs(sys_attr, type, mode)                    \
 717 do {                                                                    \
 718         sys_attr.owner = from_kuid(current_user_ns(), current_fsuid()); \
 719         sys_attr.group = from_kgid(current_user_ns(), current_fsgid()); \
 720         sys_attr.size = 0;                                              \
 721         sys_attr.perms = ORANGEFS_util_translate_mode(mode);            \
 722         sys_attr.objtype = type;                                        \
 723         sys_attr.mask = ORANGEFS_ATTR_SYS_ALL_SETABLE;                  \
 724 } while (0)
 725
 726 #define orangefs_inode_lock(__i)  mutex_lock(&(__i)->i_mutex)
 727
 728 #define orangefs_inode_unlock(__i) mutex_unlock(&(__i)->i_mutex)
 729
 730 static inline void orangefs_i_size_write(struct inode *inode, loff_t i_size)
 731 {
 732 #if BITS_PER_LONG == 32 && defined(CONFIG_SMP)
 733         orangefs_inode_lock(inode);
 734 #endif
 735         i_size_write(inode, i_size);
 736 #if BITS_PER_LONG == 32 && defined(CONFIG_SMP)
 737         orangefs_inode_unlock(inode);
 738 #endif
 739 }
 740
 741 static inline unsigned int diff(struct timeval *end, struct timeval *begin)
 742 {
 743         if (end->tv_usec < begin->tv_usec) {
 744                 end->tv_usec += 1000000;
 745                 end->tv_sec--;
 746         }
 747         end->tv_sec -= begin->tv_sec;
 748         end->tv_usec -= begin->tv_usec;
 749         return (end->tv_sec * 1000000) + end->tv_usec;
 750 }
 751
 752 #endif /* __ORANGEFSKERNEL_H */