fs/orangefs/pvfs2-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 PVFS2 Linux kernel support allows PVFS2 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 PVFS2 Linux kernel support.
  16  */
  17
  18 #ifndef __PVFS2KERNEL_H
  19 #define __PVFS2KERNEL_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 "pvfs2-dev-proto.h"
  57
  58 #ifdef PVFS2_KERNEL_DEBUG
  59 #define PVFS2_DEFAULT_OP_TIMEOUT_SECS       10
  60 #else
  61 #define PVFS2_DEFAULT_OP_TIMEOUT_SECS       20
  62 #endif
  63
  64 #define PVFS2_BUFMAP_WAIT_TIMEOUT_SECS      30
  65
  66 #define PVFS2_DEFAULT_SLOT_TIMEOUT_SECS     900 /* 15 minutes */
  67
  68 #define PVFS2_REQDEVICE_NAME          "pvfs2-req"
  69
  70 #define PVFS2_DEVREQ_MAGIC             0x20030529
  71 #define PVFS2_LINK_MAX                 0x000000FF
  72 #define PVFS2_PURGE_RETRY_COUNT        0x00000005
  73 #define PVFS2_SEEK_END                 0x00000002
  74 #define PVFS2_MAX_NUM_OPTIONS          0x00000004
  75 #define PVFS2_MAX_MOUNT_OPT_LEN        0x00000080
  76 #define PVFS2_MAX_FSKEY_LEN            64
  77
  78 #define MAX_DEV_REQ_UPSIZE (2*sizeof(__s32) +   \
  79 sizeof(__u64) + sizeof(struct pvfs2_upcall_s))
  80 #define MAX_DEV_REQ_DOWNSIZE (2*sizeof(__s32) + \
  81 sizeof(__u64) + sizeof(struct pvfs2_downcall_s))
  82
  83 #define BITS_PER_LONG_DIV_8 (BITS_PER_LONG >> 3)
  84
  85 /* borrowed from irda.h */
  86 #ifndef MSECS_TO_JIFFIES
  87 #define MSECS_TO_JIFFIES(ms) (((ms)*HZ+999)/1000)
  88 #endif
  89
  90 #define MAX_ALIGNED_DEV_REQ_UPSIZE                              \
  91                 (MAX_DEV_REQ_UPSIZE +                           \
  92                         ((((MAX_DEV_REQ_UPSIZE /                \
  93                                 (BITS_PER_LONG_DIV_8)) *        \
  94                                 (BITS_PER_LONG_DIV_8)) +        \
  95                             (BITS_PER_LONG_DIV_8)) -            \
  96                         MAX_DEV_REQ_UPSIZE))
  97
  98 #define MAX_ALIGNED_DEV_REQ_DOWNSIZE                            \
  99                 (MAX_DEV_REQ_DOWNSIZE +                         \
 100                         ((((MAX_DEV_REQ_DOWNSIZE /              \
 101                                 (BITS_PER_LONG_DIV_8)) *        \
 102                                 (BITS_PER_LONG_DIV_8)) +        \
 103                             (BITS_PER_LONG_DIV_8)) -            \
 104                         MAX_DEV_REQ_DOWNSIZE))
 105
 106 /*
 107  * valid pvfs2 kernel operation states
 108  *
 109  * unknown  - op was just initialized
 110  * waiting  - op is on request_list (upward bound)
 111  * inprogr  - op is in progress (waiting for downcall)
 112  * serviced - op has matching downcall; ok
 113  * purged   - op has to start a timer since client-core
 114  *            exited uncleanly before servicing op
 115  */
 116 enum pvfs2_vfs_op_states {
 117         OP_VFS_STATE_UNKNOWN = 0,
 118         OP_VFS_STATE_WAITING = 1,
 119         OP_VFS_STATE_INPROGR = 2,
 120         OP_VFS_STATE_SERVICED = 4,
 121         OP_VFS_STATE_PURGED = 8,
 122 };
 123
 124 #define set_op_state_waiting(op)     ((op)->op_state = OP_VFS_STATE_WAITING)
 125 #define set_op_state_inprogress(op)  ((op)->op_state = OP_VFS_STATE_INPROGR)
 126 #define set_op_state_serviced(op)    ((op)->op_state = OP_VFS_STATE_SERVICED)
 127 #define set_op_state_purged(op)      ((op)->op_state |= OP_VFS_STATE_PURGED)
 128
 129 #define op_state_waiting(op)     ((op)->op_state & OP_VFS_STATE_WAITING)
 130 #define op_state_in_progress(op) ((op)->op_state & OP_VFS_STATE_INPROGR)
 131 #define op_state_serviced(op)    ((op)->op_state & OP_VFS_STATE_SERVICED)
 132 #define op_state_purged(op)      ((op)->op_state & OP_VFS_STATE_PURGED)
 133
 134 #define get_op(op)                                      \
 135         do {                                            \
 136                 atomic_inc(&(op)->aio_ref_count);       \
 137                 gossip_debug(GOSSIP_DEV_DEBUG,  \
 138                         "(get) Alloced OP (%p:%llu)\n", \
 139                         op,                             \
 140                         llu((op)->tag));                \
 141         } while (0)
 142
 143 #define put_op(op)                                                      \
 144         do {                                                            \
 145                 if (atomic_sub_and_test(1, &(op)->aio_ref_count) == 1) {  \
 146                         gossip_debug(GOSSIP_DEV_DEBUG,          \
 147                                 "(put) Releasing OP (%p:%llu)\n",       \
 148                                 op,                                     \
 149                                 llu((op)->tag));                        \
 150                         op_release(op);                                 \
 151                         }                                               \
 152         } while (0)
 153
 154 #define op_wait(op) (atomic_read(&(op)->aio_ref_count) <= 2 ? 0 : 1)
 155
 156 /*
 157  * Defines for controlling whether I/O upcalls are for async or sync operations
 158  */
 159 enum PVFS_async_io_type {
 160         PVFS_VFS_SYNC_IO = 0,
 161         PVFS_VFS_ASYNC_IO = 1,
 162 };
 163
 164 /*
 165  * An array of client_debug_mask will be built to hold debug keyword/mask
 166  * values fetched from userspace.
 167  */
 168 struct client_debug_mask {
 169         char *keyword;
 170         __u64 mask1;
 171         __u64 mask2;
 172 };
 173
 174 /*
 175  * pvfs2 kernel memory related flags
 176  */
 177
 178 #if ((defined PVFS2_KERNEL_DEBUG) && (defined CONFIG_DEBUG_SLAB))
 179 #define PVFS2_CACHE_CREATE_FLAGS SLAB_RED_ZONE
 180 #else
 181 #define PVFS2_CACHE_CREATE_FLAGS 0
 182 #endif /* ((defined PVFS2_KERNEL_DEBUG) && (defined CONFIG_DEBUG_SLAB)) */
 183
 184 #define PVFS2_CACHE_ALLOC_FLAGS (GFP_KERNEL)
 185 #define PVFS2_GFP_FLAGS (GFP_KERNEL)
 186 #define PVFS2_BUFMAP_GFP_FLAGS (GFP_KERNEL)
 187
 188 /* pvfs2 xattr and acl related defines */
 189 #define PVFS2_XATTR_INDEX_POSIX_ACL_ACCESS  1
 190 #define PVFS2_XATTR_INDEX_POSIX_ACL_DEFAULT 2
 191 #define PVFS2_XATTR_INDEX_TRUSTED           3
 192 #define PVFS2_XATTR_INDEX_DEFAULT           4
 193
 194 #if 0
 195 #ifndef POSIX_ACL_XATTR_ACCESS
 196 #define POSIX_ACL_XATTR_ACCESS  "system.posix_acl_access"
 197 #endif
 198 #ifndef POSIX_ACL_XATTR_DEFAULT
 199 #define POSIX_ACL_XATTR_DEFAULT "system.posix_acl_default"
 200 #endif
 201 #endif
 202
 203 #define PVFS2_XATTR_NAME_ACL_ACCESS  POSIX_ACL_XATTR_ACCESS
 204 #define PVFS2_XATTR_NAME_ACL_DEFAULT POSIX_ACL_XATTR_DEFAULT
 205 #define PVFS2_XATTR_NAME_TRUSTED_PREFIX "trusted."
 206 #define PVFS2_XATTR_NAME_DEFAULT_PREFIX ""
 207
 208 /* these functions are defined in pvfs2-utils.c */
 209 int orangefs_prepare_cdm_array(char *debug_array_string);
 210 int orangefs_prepare_debugfs_help_string(int);
 211
 212 /* defined in pvfs2-debugfs.c */
 213 int pvfs2_client_debug_init(void);
 214
 215 void debug_string_to_mask(char *, void *, int);
 216 void do_c_mask(int, char *, struct client_debug_mask **);
 217 void do_k_mask(int, char *, __u64 **);
 218
 219 void debug_mask_to_string(void *, int);
 220 void do_k_string(void *, int);
 221 void do_c_string(void *, int);
 222 int check_amalgam_keyword(void *, int);
 223 int keyword_is_amalgam(char *);
 224
 225 /*these variables are defined in pvfs2-mod.c */
 226 extern char kernel_debug_string[PVFS2_MAX_DEBUG_STRING_LEN];
 227 extern char client_debug_string[PVFS2_MAX_DEBUG_STRING_LEN];
 228 extern char client_debug_array_string[PVFS2_MAX_DEBUG_STRING_LEN];
 229 extern unsigned int kernel_mask_set_mod_init;
 230
 231 extern int pvfs2_init_acl(struct inode *inode, struct inode *dir);
 232 extern const struct xattr_handler *pvfs2_xattr_handlers[];
 233
 234 extern struct posix_acl *pvfs2_get_acl(struct inode *inode, int type);
 235 extern int pvfs2_set_acl(struct inode *inode, struct posix_acl *acl, int type);
 236
 237 /*
 238  * Redefine xtvec structure so that we could move helper functions out of
 239  * the define
 240  */
 241 struct xtvec {
 242         __kernel_off_t xtv_off;         /* must be off_t */
 243         __kernel_size_t xtv_len;        /* must be size_t */
 244 };
 245
 246 /*
 247  * pvfs2 data structures
 248  */
 249 struct pvfs2_kernel_op_s {
 250         enum pvfs2_vfs_op_states op_state;
 251         __u64 tag;
 252
 253         /*
 254          * Set uses_shared_memory to 1 if this operation uses shared memory.
 255          * If true, then a retry on the op must also get a new shared memory
 256          * buffer and re-populate it.
 257          */
 258         int uses_shared_memory;
 259
 260         struct pvfs2_upcall_s upcall;
 261         struct pvfs2_downcall_s downcall;
 262
 263         wait_queue_head_t waitq;
 264         spinlock_t lock;
 265
 266         int io_completed;
 267         wait_queue_head_t io_completion_waitq;
 268
 269         /* VFS aio fields */
 270
 271         /* used by the async I/O code to stash the pvfs2_kiocb_s structure */
 272         void *priv;
 273
 274         /* used again for the async I/O code for deallocation */
 275         atomic_t aio_ref_count;
 276
 277         int attempts;
 278
 279         struct list_head list;
 280 };
 281
 282 /* per inode private pvfs2 info */
 283 struct pvfs2_inode_s {
 284         struct pvfs2_object_kref refn;
 285         char link_target[PVFS_NAME_MAX];
 286         __s64 blksize;
 287         /*
 288          * Reading/Writing Extended attributes need to acquire the appropriate
 289          * reader/writer semaphore on the pvfs2_inode_s structure.
 290          */
 291         struct rw_semaphore xattr_sem;
 292
 293         struct inode vfs_inode;
 294         sector_t last_failed_block_index_read;
 295
 296         /*
 297          * State of in-memory attributes not yet flushed to disk associated
 298          * with this object
 299          */
 300         unsigned long pinode_flags;
 301
 302         /* All allocated pvfs2_inode_s objects are chained to a list */
 303         struct list_head list;
 304 };
 305
 306 #define P_ATIME_FLAG 0
 307 #define P_MTIME_FLAG 1
 308 #define P_CTIME_FLAG 2
 309 #define P_MODE_FLAG  3
 310
 311 #define ClearAtimeFlag(pinode) clear_bit(P_ATIME_FLAG, &(pinode)->pinode_flags)
 312 #define SetAtimeFlag(pinode)   set_bit(P_ATIME_FLAG, &(pinode)->pinode_flags)
 313 #define AtimeFlag(pinode)      test_bit(P_ATIME_FLAG, &(pinode)->pinode_flags)
 314
 315 #define ClearMtimeFlag(pinode) clear_bit(P_MTIME_FLAG, &(pinode)->pinode_flags)
 316 #define SetMtimeFlag(pinode)   set_bit(P_MTIME_FLAG, &(pinode)->pinode_flags)
 317 #define MtimeFlag(pinode)      test_bit(P_MTIME_FLAG, &(pinode)->pinode_flags)
 318
 319 #define ClearCtimeFlag(pinode) clear_bit(P_CTIME_FLAG, &(pinode)->pinode_flags)
 320 #define SetCtimeFlag(pinode)   set_bit(P_CTIME_FLAG, &(pinode)->pinode_flags)
 321 #define CtimeFlag(pinode)      test_bit(P_CTIME_FLAG, &(pinode)->pinode_flags)
 322
 323 #define ClearModeFlag(pinode) clear_bit(P_MODE_FLAG, &(pinode)->pinode_flags)
 324 #define SetModeFlag(pinode)   set_bit(P_MODE_FLAG, &(pinode)->pinode_flags)
 325 #define ModeFlag(pinode)      test_bit(P_MODE_FLAG, &(pinode)->pinode_flags)
 326
 327 /* per superblock private pvfs2 info */
 328 struct pvfs2_sb_info_s {
 329         struct pvfs2_khandle root_khandle;
 330         __s32 fs_id;
 331         int id;
 332         int flags;
 333 #define PVFS2_OPT_INTR          0x01
 334 #define PVFS2_OPT_LOCAL_LOCK    0x02
 335         char devname[PVFS_MAX_SERVER_ADDR_LEN];
 336         struct super_block *sb;
 337         int mount_pending;
 338         struct list_head list;
 339 };
 340
 341 /*
 342  * structure that holds the state of any async I/O operation issued
 343  * through the VFS. Needed especially to handle cancellation requests
 344  * or even completion notification so that the VFS client-side daemon
 345  * can free up its vfs_request slots.
 346  */
 347 struct pvfs2_kiocb_s {
 348         /* the pointer to the task that initiated the AIO */
 349         struct task_struct *tsk;
 350
 351         /* pointer to the kiocb that kicked this operation */
 352         struct kiocb *kiocb;
 353
 354         /* buffer index that was used for the I/O */
 355         struct pvfs2_bufmap *bufmap;
 356         int buffer_index;
 357
 358         /* pvfs2 kernel operation type */
 359         struct pvfs2_kernel_op_s *op;
 360
 361         /* The user space buffers from/to which I/O is being staged */
 362         struct iovec *iov;
 363
 364         /* number of elements in the iovector */
 365         unsigned long nr_segs;
 366
 367         /* set to indicate the type of the operation */
 368         int rw;
 369
 370         /* file offset */
 371         loff_t offset;
 372
 373         /* and the count in bytes */
 374         size_t bytes_to_be_copied;
 375
 376         ssize_t bytes_copied;
 377         int needs_cleanup;
 378 };
 379
 380 struct pvfs2_stats {
 381         unsigned long cache_hits;
 382         unsigned long cache_misses;
 383         unsigned long reads;
 384         unsigned long writes;
 385 };
 386
 387 extern struct pvfs2_stats g_pvfs2_stats;
 388
 389 /*
 390  * NOTE: See Documentation/filesystems/porting for information
 391  * on implementing FOO_I and properly accessing fs private data
 392  */
 393 static inline struct pvfs2_inode_s *PVFS2_I(struct inode *inode)
 394 {
 395         return container_of(inode, struct pvfs2_inode_s, vfs_inode);
 396 }
 397
 398 static inline struct pvfs2_sb_info_s *PVFS2_SB(struct super_block *sb)
 399 {
 400         return (struct pvfs2_sb_info_s *) sb->s_fs_info;
 401 }
 402
 403 /* ino_t descends from "unsigned long", 8 bytes, 64 bits. */
 404 static inline ino_t pvfs2_khandle_to_ino(struct pvfs2_khandle *khandle)
 405 {
 406         union {
 407                 unsigned char u[8];
 408                 __u64 ino;
 409         } ihandle;
 410
 411         ihandle.u[0] = khandle->u[0] ^ khandle->u[4];
 412         ihandle.u[1] = khandle->u[1] ^ khandle->u[5];
 413         ihandle.u[2] = khandle->u[2] ^ khandle->u[6];
 414         ihandle.u[3] = khandle->u[3] ^ khandle->u[7];
 415         ihandle.u[4] = khandle->u[12] ^ khandle->u[8];
 416         ihandle.u[5] = khandle->u[13] ^ khandle->u[9];
 417         ihandle.u[6] = khandle->u[14] ^ khandle->u[10];
 418         ihandle.u[7] = khandle->u[15] ^ khandle->u[11];
 419
 420         return ihandle.ino;
 421 }
 422
 423 static inline struct pvfs2_khandle *get_khandle_from_ino(struct inode *inode)
 424 {
 425         return &(PVFS2_I(inode)->refn.khandle);
 426 }
 427
 428 static inline __s32 get_fsid_from_ino(struct inode *inode)
 429 {
 430         return PVFS2_I(inode)->refn.fs_id;
 431 }
 432
 433 static inline ino_t get_ino_from_khandle(struct inode *inode)
 434 {
 435         struct pvfs2_khandle *khandle;
 436         ino_t ino;
 437
 438         khandle = get_khandle_from_ino(inode);
 439         ino = pvfs2_khandle_to_ino(khandle);
 440         return ino;
 441 }
 442
 443 static inline ino_t get_parent_ino_from_dentry(struct dentry *dentry)
 444 {
 445         return get_ino_from_khandle(dentry->d_parent->d_inode);
 446 }
 447
 448 static inline int is_root_handle(struct inode *inode)
 449 {
 450         gossip_debug(GOSSIP_DCACHE_DEBUG,
 451                      "%s: root handle: %pU, this handle: %pU:\n",
 452                      __func__,
 453                      &PVFS2_SB(inode->i_sb)->root_khandle,
 454                      get_khandle_from_ino(inode));
 455
 456         if (PVFS_khandle_cmp(&(PVFS2_SB(inode->i_sb)->root_khandle),
 457                              get_khandle_from_ino(inode)))
 458                 return 0;
 459         else
 460                 return 1;
 461 }
 462
 463 static inline int match_handle(struct pvfs2_khandle resp_handle,
 464                                struct inode *inode)
 465 {
 466         gossip_debug(GOSSIP_DCACHE_DEBUG,
 467                      "%s: one handle: %pU, another handle:%pU:\n",
 468                      __func__,
 469                      &resp_handle,
 470                      get_khandle_from_ino(inode));
 471
 472         if (PVFS_khandle_cmp(&resp_handle, get_khandle_from_ino(inode)))
 473                 return 0;
 474         else
 475                 return 1;
 476 }
 477
 478 /*
 479  * defined in pvfs2-cache.c
 480  */
 481 int op_cache_initialize(void);
 482 int op_cache_finalize(void);
 483 struct pvfs2_kernel_op_s *op_alloc(__s32 type);
 484 char *get_opname_string(struct pvfs2_kernel_op_s *new_op);
 485 void op_release(struct pvfs2_kernel_op_s *op);
 486
 487 int dev_req_cache_initialize(void);
 488 int dev_req_cache_finalize(void);
 489 void *dev_req_alloc(void);
 490 void dev_req_release(void *);
 491
 492 int pvfs2_inode_cache_initialize(void);
 493 int pvfs2_inode_cache_finalize(void);
 494
 495 int kiocb_cache_initialize(void);
 496 int kiocb_cache_finalize(void);
 497 struct pvfs2_kiocb_s *kiocb_alloc(void);
 498 void kiocb_release(struct pvfs2_kiocb_s *ptr);
 499
 500 /*
 501  * defined in pvfs2-mod.c
 502  */
 503 void purge_inprogress_ops(void);
 504
 505 /*
 506  * defined in waitqueue.c
 507  */
 508 int wait_for_matching_downcall(struct pvfs2_kernel_op_s *op);
 509 int wait_for_cancellation_downcall(struct pvfs2_kernel_op_s *op);
 510 void pvfs2_clean_up_interrupted_operation(struct pvfs2_kernel_op_s *op);
 511 void purge_waiting_ops(void);
 512
 513 /*
 514  * defined in super.c
 515  */
 516 struct dentry *pvfs2_mount(struct file_system_type *fst,
 517                            int flags,
 518                            const char *devname,
 519                            void *data);
 520
 521 void pvfs2_kill_sb(struct super_block *sb);
 522 int pvfs2_remount(struct super_block *sb);
 523
 524 int fsid_key_table_initialize(void);
 525 void fsid_key_table_finalize(void);
 526
 527 /*
 528  * defined in inode.c
 529  */
 530 __u32 convert_to_pvfs2_mask(unsigned long lite_mask);
 531 struct inode *pvfs2_new_inode(struct super_block *sb,
 532                               struct inode *dir,
 533                               int mode,
 534                               dev_t dev,
 535                               struct pvfs2_object_kref *ref);
 536
 537 int pvfs2_setattr(struct dentry *dentry, struct iattr *iattr);
 538
 539 int pvfs2_getattr(struct vfsmount *mnt,
 540                   struct dentry *dentry,
 541                   struct kstat *kstat);
 542
 543 /*
 544  * defined in xattr.c
 545  */
 546 int pvfs2_setxattr(struct dentry *dentry,
 547                    const char *name,
 548                    const void *value,
 549                    size_t size,
 550                    int flags);
 551
 552 ssize_t pvfs2_getxattr(struct dentry *dentry,
 553                        const char *name,
 554                        void *buffer,
 555                        size_t size);
 556
 557 ssize_t pvfs2_listxattr(struct dentry *dentry, char *buffer, size_t size);
 558
 559 /*
 560  * defined in namei.c
 561  */
 562 struct inode *pvfs2_iget(struct super_block *sb,
 563                          struct pvfs2_object_kref *ref);
 564
 565 ssize_t pvfs2_inode_read(struct inode *inode,
 566                          struct iov_iter *iter,
 567                          loff_t *offset,
 568                          loff_t readahead_size);
 569
 570 /*
 571  * defined in devpvfs2-req.c
 572  */
 573 int pvfs2_dev_init(void);
 574 void pvfs2_dev_cleanup(void);
 575 int is_daemon_in_service(void);
 576 int fs_mount_pending(__s32 fsid);
 577
 578 /*
 579  * defined in pvfs2-utils.c
 580  */
 581 __s32 fsid_of_op(struct pvfs2_kernel_op_s *op);
 582
 583 int pvfs2_flush_inode(struct inode *inode);
 584
 585 ssize_t pvfs2_inode_getxattr(struct inode *inode,
 586                              const char *prefix,
 587                              const char *name,
 588                              void *buffer,
 589                              size_t size);
 590
 591 int pvfs2_inode_setxattr(struct inode *inode,
 592                          const char *prefix,
 593                          const char *name,
 594                          const void *value,
 595                          size_t size,
 596                          int flags);
 597
 598 int pvfs2_inode_getattr(struct inode *inode, __u32 mask);
 599
 600 int pvfs2_inode_setattr(struct inode *inode, struct iattr *iattr);
 601
 602 void pvfs2_op_initialize(struct pvfs2_kernel_op_s *op);
 603
 604 void pvfs2_make_bad_inode(struct inode *inode);
 605
 606 void block_signals(sigset_t *);
 607
 608 void set_signals(sigset_t *);
 609
 610 int pvfs2_unmount_sb(struct super_block *sb);
 611
 612 int pvfs2_cancel_op_in_progress(__u64 tag);
 613
 614 static inline __u64 pvfs2_convert_time_field(const struct timespec *ts)
 615 {
 616         return (__u64)ts->tv_sec;
 617 }
 618
 619 int pvfs2_normalize_to_errno(__s32 error_code);
 620
 621 extern struct mutex devreq_mutex;
 622 extern struct mutex request_mutex;
 623 extern int debug;
 624 extern int op_timeout_secs;
 625 extern int slot_timeout_secs;
 626 extern struct list_head pvfs2_superblocks;
 627 extern spinlock_t pvfs2_superblocks_lock;
 628 extern struct list_head pvfs2_request_list;
 629 extern spinlock_t pvfs2_request_list_lock;
 630 extern wait_queue_head_t pvfs2_request_list_waitq;
 631 extern struct list_head *htable_ops_in_progress;
 632 extern spinlock_t htable_ops_in_progress_lock;
 633 extern int hash_table_size;
 634
 635 extern const struct address_space_operations pvfs2_address_operations;
 636 extern struct backing_dev_info pvfs2_backing_dev_info;
 637 extern struct inode_operations pvfs2_file_inode_operations;
 638 extern const struct file_operations pvfs2_file_operations;
 639 extern struct inode_operations pvfs2_symlink_inode_operations;
 640 extern struct inode_operations pvfs2_dir_inode_operations;
 641 extern const struct file_operations pvfs2_dir_operations;
 642 extern const struct dentry_operations pvfs2_dentry_operations;
 643 extern const struct file_operations pvfs2_devreq_file_operations;
 644
 645 extern wait_queue_head_t pvfs2_bufmap_init_waitq;
 646
 647 /*
 648  * misc convenience macros
 649  */
 650 #define add_op_to_request_list(op)                              \
 651 do {                                                            \
 652         spin_lock(&pvfs2_request_list_lock);                    \
 653         spin_lock(&op->lock);                                   \
 654         set_op_state_waiting(op);                               \
 655         list_add_tail(&op->list, &pvfs2_request_list);          \
 656         spin_unlock(&pvfs2_request_list_lock);                  \
 657         spin_unlock(&op->lock);                                 \
 658         wake_up_interruptible(&pvfs2_request_list_waitq);       \
 659 } while (0)
 660
 661 #define add_priority_op_to_request_list(op)                             \
 662         do {                                                            \
 663                 spin_lock(&pvfs2_request_list_lock);                    \
 664                 spin_lock(&op->lock);                                   \
 665                 set_op_state_waiting(op);                               \
 666                                                                         \
 667                 list_add(&op->list, &pvfs2_request_list);               \
 668                 spin_unlock(&pvfs2_request_list_lock);                  \
 669                 spin_unlock(&op->lock);                                 \
 670                 wake_up_interruptible(&pvfs2_request_list_waitq);       \
 671 } while (0)
 672
 673 #define remove_op_from_request_list(op)                                 \
 674         do {                                                            \
 675                 struct list_head *tmp = NULL;                           \
 676                 struct list_head *tmp_safe = NULL;                      \
 677                 struct pvfs2_kernel_op_s *tmp_op = NULL;                \
 678                                                                         \
 679                 spin_lock(&pvfs2_request_list_lock);                    \
 680                 list_for_each_safe(tmp, tmp_safe, &pvfs2_request_list) { \
 681                         tmp_op = list_entry(tmp,                        \
 682                                             struct pvfs2_kernel_op_s,   \
 683                                             list);                      \
 684                         if (tmp_op && (tmp_op == op)) {                 \
 685                                 list_del(&tmp_op->list);                \
 686                                 break;                                  \
 687                         }                                               \
 688                 }                                                       \
 689                 spin_unlock(&pvfs2_request_list_lock);                  \
 690         } while (0)
 691
 692 #define PVFS2_OP_INTERRUPTIBLE 1   /* service_operation() is interruptible */
 693 #define PVFS2_OP_PRIORITY      2   /* service_operation() is high priority */
 694 #define PVFS2_OP_CANCELLATION  4   /* this is a cancellation */
 695 #define PVFS2_OP_NO_SEMAPHORE  8   /* don't acquire semaphore */
 696 #define PVFS2_OP_ASYNC         16  /* Queue it, but don't wait */
 697
 698 int service_operation(struct pvfs2_kernel_op_s *op,
 699                       const char *op_name,
 700                       int flags);
 701
 702 /*
 703  * handles two possible error cases, depending on context.
 704  *
 705  * by design, our vfs i/o errors need to be handled in one of two ways,
 706  * depending on where the error occured.
 707  *
 708  * if the error happens in the waitqueue code because we either timed
 709  * out or a signal was raised while waiting, we need to cancel the
 710  * userspace i/o operation and free the op manually.  this is done to
 711  * avoid having the device start writing application data to our shared
 712  * bufmap pages without us expecting it.
 713  *
 714  * FIXME: POSSIBLE OPTIMIZATION:
 715  * However, if we timed out or if we got a signal AND our upcall was never
 716  * picked off the queue (i.e. we were in OP_VFS_STATE_WAITING), then we don't
 717  * need to send a cancellation upcall. The way we can handle this is
 718  * set error_exit to 2 in such cases and 1 whenever cancellation has to be
 719  * sent and have handle_error
 720  * take care of this situation as well..
 721  *
 722  * if a pvfs2 sysint level error occured and i/o has been completed,
 723  * there is no need to cancel the operation, as the user has finished
 724  * using the bufmap page and so there is no danger in this case.  in
 725  * this case, we wake up the device normally so that it may free the
 726  * op, as normal.
 727  *
 728  * note the only reason this is a macro is because both read and write
 729  * cases need the exact same handling code.
 730  */
 731 #define handle_io_error()                                       \
 732 do {                                                            \
 733         if (!op_state_serviced(new_op)) {                       \
 734                 pvfs2_cancel_op_in_progress(new_op->tag);       \
 735                 op_release(new_op);                             \
 736         } else {                                                \
 737                 wake_up_daemon_for_return(new_op);              \
 738         }                                                       \
 739         new_op = NULL;                                          \
 740         pvfs_bufmap_put(bufmap, buffer_index);                          \
 741         buffer_index = -1;                                      \
 742 } while (0)
 743
 744 #define get_interruptible_flag(inode) \
 745         ((PVFS2_SB(inode->i_sb)->flags & PVFS2_OPT_INTR) ? \
 746                 PVFS2_OP_INTERRUPTIBLE : 0)
 747
 748 #define add_pvfs2_sb(sb)                                                \
 749 do {                                                                    \
 750         gossip_debug(GOSSIP_SUPER_DEBUG,                                \
 751                      "Adding SB %p to pvfs2 superblocks\n",             \
 752                      PVFS2_SB(sb));                                     \
 753         spin_lock(&pvfs2_superblocks_lock);                             \
 754         list_add_tail(&PVFS2_SB(sb)->list, &pvfs2_superblocks);         \
 755         spin_unlock(&pvfs2_superblocks_lock); \
 756 } while (0)
 757
 758 #define remove_pvfs2_sb(sb)                                             \
 759 do {                                                                    \
 760         struct list_head *tmp = NULL;                                   \
 761         struct list_head *tmp_safe = NULL;                              \
 762         struct pvfs2_sb_info_s *pvfs2_sb = NULL;                        \
 763                                                                         \
 764         spin_lock(&pvfs2_superblocks_lock);                             \
 765         list_for_each_safe(tmp, tmp_safe, &pvfs2_superblocks) {         \
 766                 pvfs2_sb = list_entry(tmp,                              \
 767                                       struct pvfs2_sb_info_s,           \
 768                                       list);                            \
 769                 if (pvfs2_sb && (pvfs2_sb->sb == sb)) {                 \
 770                         gossip_debug(GOSSIP_SUPER_DEBUG,                \
 771                             "Removing SB %p from pvfs2 superblocks\n",  \
 772                         pvfs2_sb);                                      \
 773                         list_del(&pvfs2_sb->list);                      \
 774                         break;                                          \
 775                 }                                                       \
 776         }                                                               \
 777         spin_unlock(&pvfs2_superblocks_lock);                           \
 778 } while (0)
 779
 780 #define pvfs2_lock_inode(inode) spin_lock(&inode->i_lock)
 781 #define pvfs2_unlock_inode(inode) spin_unlock(&inode->i_lock)
 782
 783 #define fill_default_sys_attrs(sys_attr, type, mode)                    \
 784 do {                                                                    \
 785         sys_attr.owner = from_kuid(current_user_ns(), current_fsuid()); \
 786         sys_attr.group = from_kgid(current_user_ns(), current_fsgid()); \
 787         sys_attr.size = 0;                                              \
 788         sys_attr.perms = PVFS_util_translate_mode(mode);                \
 789         sys_attr.objtype = type;                                        \
 790         sys_attr.mask = PVFS_ATTR_SYS_ALL_SETABLE;                      \
 791 } while (0)
 792
 793 #define pvfs2_inode_lock(__i)  mutex_lock(&(__i)->i_mutex)
 794
 795 #define pvfs2_inode_unlock(__i) mutex_unlock(&(__i)->i_mutex)
 796
 797 static inline void pvfs2_i_size_write(struct inode *inode, loff_t i_size)
 798 {
 799 #if BITS_PER_LONG == 32 && defined(CONFIG_SMP)
 800         pvfs2_inode_lock(inode);
 801 #endif
 802         i_size_write(inode, i_size);
 803 #if BITS_PER_LONG == 32 && defined(CONFIG_SMP)
 804         pvfs2_inode_unlock(inode);
 805 #endif
 806 }
 807
 808 static inline unsigned int diff(struct timeval *end, struct timeval *begin)
 809 {
 810         if (end->tv_usec < begin->tv_usec) {
 811                 end->tv_usec += 1000000;
 812                 end->tv_sec--;
 813         }
 814         end->tv_sec -= begin->tv_sec;
 815         end->tv_usec -= begin->tv_usec;
 816         return (end->tv_sec * 1000000) + end->tv_usec;
 817 }
 818
 819 #endif /* __PVFS2KERNEL_H */