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