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