Mike Marshall | f7ab093 | 2015-07-17 10:38:11 -0400 | [diff] [blame^] | 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 | #define pvfs2_kmap(page) kmap(page) |
| 189 | #define pvfs2_kunmap(page) kunmap(page) |
| 190 | |
| 191 | /* pvfs2 xattr and acl related defines */ |
| 192 | #define PVFS2_XATTR_INDEX_POSIX_ACL_ACCESS 1 |
| 193 | #define PVFS2_XATTR_INDEX_POSIX_ACL_DEFAULT 2 |
| 194 | #define PVFS2_XATTR_INDEX_TRUSTED 3 |
| 195 | #define PVFS2_XATTR_INDEX_DEFAULT 4 |
| 196 | |
| 197 | #if 0 |
| 198 | #ifndef POSIX_ACL_XATTR_ACCESS |
| 199 | #define POSIX_ACL_XATTR_ACCESS "system.posix_acl_access" |
| 200 | #endif |
| 201 | #ifndef POSIX_ACL_XATTR_DEFAULT |
| 202 | #define POSIX_ACL_XATTR_DEFAULT "system.posix_acl_default" |
| 203 | #endif |
| 204 | #endif |
| 205 | |
| 206 | #define PVFS2_XATTR_NAME_ACL_ACCESS POSIX_ACL_XATTR_ACCESS |
| 207 | #define PVFS2_XATTR_NAME_ACL_DEFAULT POSIX_ACL_XATTR_DEFAULT |
| 208 | #define PVFS2_XATTR_NAME_TRUSTED_PREFIX "trusted." |
| 209 | #define PVFS2_XATTR_NAME_DEFAULT_PREFIX "" |
| 210 | |
| 211 | /* these functions are defined in pvfs2-utils.c */ |
| 212 | int orangefs_prepare_cdm_array(char *debug_array_string); |
| 213 | int orangefs_prepare_debugfs_help_string(int); |
| 214 | |
| 215 | /* defined in pvfs2-debugfs.c */ |
| 216 | int pvfs2_client_debug_init(void); |
| 217 | |
| 218 | void debug_string_to_mask(char *, void *, int); |
| 219 | void do_c_mask(int, char *, struct client_debug_mask **); |
| 220 | void do_k_mask(int, char *, __u64 **); |
| 221 | |
| 222 | void debug_mask_to_string(void *, int); |
| 223 | void do_k_string(void *, int); |
| 224 | void do_c_string(void *, int); |
| 225 | int check_amalgam_keyword(void *, int); |
| 226 | int keyword_is_amalgam(char *); |
| 227 | |
| 228 | /*these variables are defined in pvfs2-mod.c */ |
| 229 | extern char kernel_debug_string[PVFS2_MAX_DEBUG_STRING_LEN]; |
| 230 | extern char client_debug_string[PVFS2_MAX_DEBUG_STRING_LEN]; |
| 231 | extern char client_debug_array_string[PVFS2_MAX_DEBUG_STRING_LEN]; |
| 232 | /* HELLO |
| 233 | extern struct client_debug_mask current_client_mask; |
| 234 | */ |
| 235 | extern unsigned int kernel_mask_set_mod_init; |
| 236 | |
| 237 | extern int pvfs2_init_acl(struct inode *inode, struct inode *dir); |
| 238 | extern const struct xattr_handler *pvfs2_xattr_handlers[]; |
| 239 | |
| 240 | extern struct posix_acl *pvfs2_get_acl(struct inode *inode, int type); |
| 241 | extern int pvfs2_set_acl(struct inode *inode, struct posix_acl *acl, int type); |
| 242 | |
| 243 | int pvfs2_xattr_set_default(struct dentry *dentry, |
| 244 | const char *name, |
| 245 | const void *buffer, |
| 246 | size_t size, |
| 247 | int flags, |
| 248 | int handler_flags); |
| 249 | |
| 250 | int pvfs2_xattr_get_default(struct dentry *dentry, |
| 251 | const char *name, |
| 252 | void *buffer, |
| 253 | size_t size, |
| 254 | int handler_flags); |
| 255 | |
| 256 | /* |
| 257 | * Redefine xtvec structure so that we could move helper functions out of |
| 258 | * the define |
| 259 | */ |
| 260 | struct xtvec { |
| 261 | __kernel_off_t xtv_off; /* must be off_t */ |
| 262 | __kernel_size_t xtv_len; /* must be size_t */ |
| 263 | }; |
| 264 | |
| 265 | /* |
| 266 | * pvfs2 data structures |
| 267 | */ |
| 268 | struct pvfs2_kernel_op_s { |
| 269 | enum pvfs2_vfs_op_states op_state; |
| 270 | __u64 tag; |
| 271 | |
| 272 | /* |
| 273 | * Set uses_shared_memory to 1 if this operation uses shared memory. |
| 274 | * If true, then a retry on the op must also get a new shared memory |
| 275 | * buffer and re-populate it. |
| 276 | */ |
| 277 | int uses_shared_memory; |
| 278 | |
| 279 | struct pvfs2_upcall_s upcall; |
| 280 | struct pvfs2_downcall_s downcall; |
| 281 | |
| 282 | wait_queue_head_t waitq; |
| 283 | spinlock_t lock; |
| 284 | |
| 285 | int io_completed; |
| 286 | wait_queue_head_t io_completion_waitq; |
| 287 | |
| 288 | /* |
| 289 | * upcalls requiring variable length trailers require that this struct |
| 290 | * be in the request list even after client-core does a read() on the |
| 291 | * device to dequeue the upcall. |
| 292 | * if op_linger field goes to 0, we dequeue this op off the list. |
| 293 | * else we let it stay. What gets passed to the read() is |
| 294 | * a) if op_linger field is = 1, pvfs2_kernel_op_s itself |
| 295 | * b) else if = 0, we pass ->upcall.trailer_buf |
| 296 | * We expect to have only a single upcall trailer buffer, |
| 297 | * so we expect callers with trailers |
| 298 | * to set this field to 2 and others to set it to 1. |
| 299 | */ |
| 300 | __s32 op_linger, op_linger_tmp; |
| 301 | /* VFS aio fields */ |
| 302 | |
| 303 | /* used by the async I/O code to stash the pvfs2_kiocb_s structure */ |
| 304 | void *priv; |
| 305 | |
| 306 | /* used again for the async I/O code for deallocation */ |
| 307 | atomic_t aio_ref_count; |
| 308 | |
| 309 | int attempts; |
| 310 | |
| 311 | struct list_head list; |
| 312 | }; |
| 313 | |
| 314 | /* per inode private pvfs2 info */ |
| 315 | struct pvfs2_inode_s { |
| 316 | struct pvfs2_object_kref refn; |
| 317 | char link_target[PVFS_NAME_MAX]; |
| 318 | __s64 blksize; |
| 319 | /* |
| 320 | * Reading/Writing Extended attributes need to acquire the appropriate |
| 321 | * reader/writer semaphore on the pvfs2_inode_s structure. |
| 322 | */ |
| 323 | struct rw_semaphore xattr_sem; |
| 324 | |
| 325 | struct inode vfs_inode; |
| 326 | sector_t last_failed_block_index_read; |
| 327 | |
| 328 | /* |
| 329 | * State of in-memory attributes not yet flushed to disk associated |
| 330 | * with this object |
| 331 | */ |
| 332 | unsigned long pinode_flags; |
| 333 | |
| 334 | /* All allocated pvfs2_inode_s objects are chained to a list */ |
| 335 | struct list_head list; |
| 336 | }; |
| 337 | |
| 338 | #define P_ATIME_FLAG 0 |
| 339 | #define P_MTIME_FLAG 1 |
| 340 | #define P_CTIME_FLAG 2 |
| 341 | #define P_MODE_FLAG 3 |
| 342 | |
| 343 | #define ClearAtimeFlag(pinode) clear_bit(P_ATIME_FLAG, &(pinode)->pinode_flags) |
| 344 | #define SetAtimeFlag(pinode) set_bit(P_ATIME_FLAG, &(pinode)->pinode_flags) |
| 345 | #define AtimeFlag(pinode) test_bit(P_ATIME_FLAG, &(pinode)->pinode_flags) |
| 346 | |
| 347 | #define ClearMtimeFlag(pinode) clear_bit(P_MTIME_FLAG, &(pinode)->pinode_flags) |
| 348 | #define SetMtimeFlag(pinode) set_bit(P_MTIME_FLAG, &(pinode)->pinode_flags) |
| 349 | #define MtimeFlag(pinode) test_bit(P_MTIME_FLAG, &(pinode)->pinode_flags) |
| 350 | |
| 351 | #define ClearCtimeFlag(pinode) clear_bit(P_CTIME_FLAG, &(pinode)->pinode_flags) |
| 352 | #define SetCtimeFlag(pinode) set_bit(P_CTIME_FLAG, &(pinode)->pinode_flags) |
| 353 | #define CtimeFlag(pinode) test_bit(P_CTIME_FLAG, &(pinode)->pinode_flags) |
| 354 | |
| 355 | #define ClearModeFlag(pinode) clear_bit(P_MODE_FLAG, &(pinode)->pinode_flags) |
| 356 | #define SetModeFlag(pinode) set_bit(P_MODE_FLAG, &(pinode)->pinode_flags) |
| 357 | #define ModeFlag(pinode) test_bit(P_MODE_FLAG, &(pinode)->pinode_flags) |
| 358 | |
| 359 | /* per superblock private pvfs2 info */ |
| 360 | struct pvfs2_sb_info_s { |
| 361 | struct pvfs2_khandle root_khandle; |
| 362 | __s32 fs_id; |
| 363 | int id; |
| 364 | int flags; |
| 365 | #define PVFS2_OPT_INTR 0x01 |
| 366 | #define PVFS2_OPT_LOCAL_LOCK 0x02 |
| 367 | char devname[PVFS_MAX_SERVER_ADDR_LEN]; |
| 368 | struct super_block *sb; |
| 369 | int mount_pending; |
| 370 | struct list_head list; |
| 371 | }; |
| 372 | |
| 373 | /* |
| 374 | * a temporary structure used only for sb mount time that groups the |
| 375 | * mount time data provided along with a private superblock structure |
| 376 | * that is allocated before a 'kernel' superblock is allocated. |
| 377 | */ |
| 378 | struct pvfs2_mount_sb_info_s { |
| 379 | void *data; |
| 380 | struct pvfs2_khandle root_khandle; |
| 381 | __s32 fs_id; |
| 382 | int id; |
| 383 | }; |
| 384 | |
| 385 | /* |
| 386 | * structure that holds the state of any async I/O operation issued |
| 387 | * through the VFS. Needed especially to handle cancellation requests |
| 388 | * or even completion notification so that the VFS client-side daemon |
| 389 | * can free up its vfs_request slots. |
| 390 | */ |
| 391 | struct pvfs2_kiocb_s { |
| 392 | /* the pointer to the task that initiated the AIO */ |
| 393 | struct task_struct *tsk; |
| 394 | |
| 395 | /* pointer to the kiocb that kicked this operation */ |
| 396 | struct kiocb *kiocb; |
| 397 | |
| 398 | /* buffer index that was used for the I/O */ |
| 399 | struct pvfs2_bufmap *bufmap; |
| 400 | int buffer_index; |
| 401 | |
| 402 | /* pvfs2 kernel operation type */ |
| 403 | struct pvfs2_kernel_op_s *op; |
| 404 | |
| 405 | /* The user space buffers from/to which I/O is being staged */ |
| 406 | struct iovec *iov; |
| 407 | |
| 408 | /* number of elements in the iovector */ |
| 409 | unsigned long nr_segs; |
| 410 | |
| 411 | /* set to indicate the type of the operation */ |
| 412 | int rw; |
| 413 | |
| 414 | /* file offset */ |
| 415 | loff_t offset; |
| 416 | |
| 417 | /* and the count in bytes */ |
| 418 | size_t bytes_to_be_copied; |
| 419 | |
| 420 | ssize_t bytes_copied; |
| 421 | int needs_cleanup; |
| 422 | }; |
| 423 | |
| 424 | struct pvfs2_stats { |
| 425 | unsigned long cache_hits; |
| 426 | unsigned long cache_misses; |
| 427 | unsigned long reads; |
| 428 | unsigned long writes; |
| 429 | }; |
| 430 | |
| 431 | extern struct pvfs2_stats g_pvfs2_stats; |
| 432 | |
| 433 | /* |
| 434 | NOTE: See Documentation/filesystems/porting for information |
| 435 | on implementing FOO_I and properly accessing fs private data |
| 436 | */ |
| 437 | static inline struct pvfs2_inode_s *PVFS2_I(struct inode *inode) |
| 438 | { |
| 439 | return container_of(inode, struct pvfs2_inode_s, vfs_inode); |
| 440 | } |
| 441 | |
| 442 | static inline struct pvfs2_sb_info_s *PVFS2_SB(struct super_block *sb) |
| 443 | { |
| 444 | return (struct pvfs2_sb_info_s *) sb->s_fs_info; |
| 445 | } |
| 446 | |
| 447 | /* ino_t descends from "unsigned long", 8 bytes, 64 bits. */ |
| 448 | static inline ino_t pvfs2_khandle_to_ino(struct pvfs2_khandle *khandle) |
| 449 | { |
| 450 | union { |
| 451 | unsigned char u[8]; |
| 452 | __u64 ino; |
| 453 | } ihandle; |
| 454 | |
| 455 | ihandle.u[0] = khandle->u[0] ^ khandle->u[4]; |
| 456 | ihandle.u[1] = khandle->u[1] ^ khandle->u[5]; |
| 457 | ihandle.u[2] = khandle->u[2] ^ khandle->u[6]; |
| 458 | ihandle.u[3] = khandle->u[3] ^ khandle->u[7]; |
| 459 | ihandle.u[4] = khandle->u[12] ^ khandle->u[8]; |
| 460 | ihandle.u[5] = khandle->u[13] ^ khandle->u[9]; |
| 461 | ihandle.u[6] = khandle->u[14] ^ khandle->u[10]; |
| 462 | ihandle.u[7] = khandle->u[15] ^ khandle->u[11]; |
| 463 | |
| 464 | return ihandle.ino; |
| 465 | } |
| 466 | |
| 467 | static inline struct pvfs2_khandle *get_khandle_from_ino(struct inode *inode) |
| 468 | { |
| 469 | return &(PVFS2_I(inode)->refn.khandle); |
| 470 | } |
| 471 | |
| 472 | static inline __s32 get_fsid_from_ino(struct inode *inode) |
| 473 | { |
| 474 | return PVFS2_I(inode)->refn.fs_id; |
| 475 | } |
| 476 | |
| 477 | static inline ino_t get_ino_from_khandle(struct inode *inode) |
| 478 | { |
| 479 | struct pvfs2_khandle *khandle; |
| 480 | ino_t ino; |
| 481 | |
| 482 | khandle = get_khandle_from_ino(inode); |
| 483 | ino = pvfs2_khandle_to_ino(khandle); |
| 484 | return ino; |
| 485 | } |
| 486 | |
| 487 | static inline ino_t get_parent_ino_from_dentry(struct dentry *dentry) |
| 488 | { |
| 489 | return get_ino_from_khandle(dentry->d_parent->d_inode); |
| 490 | } |
| 491 | |
| 492 | static inline int is_root_handle(struct inode *inode) |
| 493 | { |
| 494 | gossip_debug(GOSSIP_DCACHE_DEBUG, |
| 495 | "%s: root handle: %pU, this handle: %pU:\n", |
| 496 | __func__, |
| 497 | &PVFS2_SB(inode->i_sb)->root_khandle, |
| 498 | get_khandle_from_ino(inode)); |
| 499 | |
| 500 | if (PVFS_khandle_cmp(&(PVFS2_SB(inode->i_sb)->root_khandle), |
| 501 | get_khandle_from_ino(inode))) |
| 502 | return 0; |
| 503 | else |
| 504 | return 1; |
| 505 | } |
| 506 | |
| 507 | static inline int match_handle(struct pvfs2_khandle resp_handle, |
| 508 | struct inode *inode) |
| 509 | { |
| 510 | gossip_debug(GOSSIP_DCACHE_DEBUG, |
| 511 | "%s: one handle: %pU, another handle:%pU:\n", |
| 512 | __func__, |
| 513 | &resp_handle, |
| 514 | get_khandle_from_ino(inode)); |
| 515 | |
| 516 | if (PVFS_khandle_cmp(&resp_handle, get_khandle_from_ino(inode))) |
| 517 | return 0; |
| 518 | else |
| 519 | return 1; |
| 520 | } |
| 521 | |
| 522 | /* |
| 523 | * defined in pvfs2-cache.c |
| 524 | */ |
| 525 | int op_cache_initialize(void); |
| 526 | int op_cache_finalize(void); |
| 527 | struct pvfs2_kernel_op_s *op_alloc(__s32 type); |
| 528 | struct pvfs2_kernel_op_s *op_alloc_trailer(__s32 type); |
| 529 | char *get_opname_string(struct pvfs2_kernel_op_s *new_op); |
| 530 | void op_release(struct pvfs2_kernel_op_s *op); |
| 531 | |
| 532 | int dev_req_cache_initialize(void); |
| 533 | int dev_req_cache_finalize(void); |
| 534 | void *dev_req_alloc(void); |
| 535 | void dev_req_release(void *); |
| 536 | |
| 537 | int pvfs2_inode_cache_initialize(void); |
| 538 | int pvfs2_inode_cache_finalize(void); |
| 539 | |
| 540 | int kiocb_cache_initialize(void); |
| 541 | int kiocb_cache_finalize(void); |
| 542 | struct pvfs2_kiocb_s *kiocb_alloc(void); |
| 543 | void kiocb_release(struct pvfs2_kiocb_s *ptr); |
| 544 | |
| 545 | /* |
| 546 | * defined in pvfs2-mod.c |
| 547 | */ |
| 548 | void purge_inprogress_ops(void); |
| 549 | |
| 550 | /* |
| 551 | * defined in waitqueue.c |
| 552 | */ |
| 553 | int wait_for_matching_downcall(struct pvfs2_kernel_op_s *op); |
| 554 | int wait_for_cancellation_downcall(struct pvfs2_kernel_op_s *op); |
| 555 | void pvfs2_clean_up_interrupted_operation(struct pvfs2_kernel_op_s *op); |
| 556 | void purge_waiting_ops(void); |
| 557 | |
| 558 | /* |
| 559 | * defined in super.c |
| 560 | */ |
| 561 | struct dentry *pvfs2_mount(struct file_system_type *fst, |
| 562 | int flags, |
| 563 | const char *devname, |
| 564 | void *data); |
| 565 | |
| 566 | void pvfs2_kill_sb(struct super_block *sb); |
| 567 | int pvfs2_remount(struct super_block *sb); |
| 568 | |
| 569 | int fsid_key_table_initialize(void); |
| 570 | void fsid_key_table_finalize(void); |
| 571 | |
| 572 | /* |
| 573 | * defined in inode.c |
| 574 | */ |
| 575 | __u32 convert_to_pvfs2_mask(unsigned long lite_mask); |
| 576 | struct inode *pvfs2_new_inode(struct super_block *sb, |
| 577 | struct inode *dir, |
| 578 | int mode, |
| 579 | dev_t dev, |
| 580 | struct pvfs2_object_kref *ref); |
| 581 | |
| 582 | int pvfs2_setattr(struct dentry *dentry, struct iattr *iattr); |
| 583 | |
| 584 | int pvfs2_getattr(struct vfsmount *mnt, |
| 585 | struct dentry *dentry, |
| 586 | struct kstat *kstat); |
| 587 | |
| 588 | /* |
| 589 | * defined in xattr.c |
| 590 | */ |
| 591 | int pvfs2_setxattr(struct dentry *dentry, |
| 592 | const char *name, |
| 593 | const void *value, |
| 594 | size_t size, |
| 595 | int flags); |
| 596 | |
| 597 | ssize_t pvfs2_getxattr(struct dentry *dentry, |
| 598 | const char *name, |
| 599 | void *buffer, |
| 600 | size_t size); |
| 601 | |
| 602 | ssize_t pvfs2_listxattr(struct dentry *dentry, char *buffer, size_t size); |
| 603 | |
| 604 | /* |
| 605 | * defined in namei.c |
| 606 | */ |
| 607 | struct inode *pvfs2_iget(struct super_block *sb, |
| 608 | struct pvfs2_object_kref *ref); |
| 609 | |
| 610 | ssize_t pvfs2_inode_read(struct inode *inode, |
| 611 | char *buf, |
| 612 | size_t count, |
| 613 | loff_t *offset, |
| 614 | loff_t readahead_size); |
| 615 | |
| 616 | /* |
| 617 | * defined in devpvfs2-req.c |
| 618 | */ |
| 619 | int pvfs2_dev_init(void); |
| 620 | void pvfs2_dev_cleanup(void); |
| 621 | int is_daemon_in_service(void); |
| 622 | int fs_mount_pending(__s32 fsid); |
| 623 | |
| 624 | /* |
| 625 | * defined in pvfs2-utils.c |
| 626 | */ |
| 627 | __s32 fsid_of_op(struct pvfs2_kernel_op_s *op); |
| 628 | |
| 629 | int pvfs2_flush_inode(struct inode *inode); |
| 630 | |
| 631 | ssize_t pvfs2_inode_getxattr(struct inode *inode, |
| 632 | const char *prefix, |
| 633 | const char *name, |
| 634 | void *buffer, |
| 635 | size_t size); |
| 636 | |
| 637 | int pvfs2_inode_setxattr(struct inode *inode, |
| 638 | const char *prefix, |
| 639 | const char *name, |
| 640 | const void *value, |
| 641 | size_t size, |
| 642 | int flags); |
| 643 | |
| 644 | int pvfs2_inode_getattr(struct inode *inode, __u32 mask); |
| 645 | |
| 646 | int pvfs2_inode_setattr(struct inode *inode, struct iattr *iattr); |
| 647 | |
| 648 | void pvfs2_op_initialize(struct pvfs2_kernel_op_s *op); |
| 649 | |
| 650 | void pvfs2_make_bad_inode(struct inode *inode); |
| 651 | |
| 652 | void mask_blocked_signals(sigset_t *orig_sigset); |
| 653 | |
| 654 | void unmask_blocked_signals(sigset_t *orig_sigset); |
| 655 | |
| 656 | int pvfs2_unmount_sb(struct super_block *sb); |
| 657 | |
| 658 | int pvfs2_cancel_op_in_progress(__u64 tag); |
| 659 | |
| 660 | __u64 pvfs2_convert_time_field(void *time_ptr); |
| 661 | |
| 662 | int pvfs2_normalize_to_errno(__s32 error_code); |
| 663 | |
| 664 | extern struct mutex devreq_mutex; |
| 665 | extern struct mutex request_mutex; |
| 666 | extern int debug; |
| 667 | extern int op_timeout_secs; |
| 668 | extern int slot_timeout_secs; |
| 669 | extern struct list_head pvfs2_superblocks; |
| 670 | extern spinlock_t pvfs2_superblocks_lock; |
| 671 | extern struct list_head pvfs2_request_list; |
| 672 | extern spinlock_t pvfs2_request_list_lock; |
| 673 | extern wait_queue_head_t pvfs2_request_list_waitq; |
| 674 | extern struct list_head *htable_ops_in_progress; |
| 675 | extern spinlock_t htable_ops_in_progress_lock; |
| 676 | extern int hash_table_size; |
| 677 | |
| 678 | extern const struct address_space_operations pvfs2_address_operations; |
| 679 | extern struct backing_dev_info pvfs2_backing_dev_info; |
| 680 | extern struct inode_operations pvfs2_file_inode_operations; |
| 681 | extern const struct file_operations pvfs2_file_operations; |
| 682 | extern struct inode_operations pvfs2_symlink_inode_operations; |
| 683 | extern struct inode_operations pvfs2_dir_inode_operations; |
| 684 | extern const struct file_operations pvfs2_dir_operations; |
| 685 | extern const struct dentry_operations pvfs2_dentry_operations; |
| 686 | extern const struct file_operations pvfs2_devreq_file_operations; |
| 687 | |
| 688 | extern wait_queue_head_t pvfs2_bufmap_init_waitq; |
| 689 | |
| 690 | /* |
| 691 | * misc convenience macros |
| 692 | */ |
| 693 | #define add_op_to_request_list(op) \ |
| 694 | do { \ |
| 695 | spin_lock(&pvfs2_request_list_lock); \ |
| 696 | spin_lock(&op->lock); \ |
| 697 | set_op_state_waiting(op); \ |
| 698 | list_add_tail(&op->list, &pvfs2_request_list); \ |
| 699 | spin_unlock(&pvfs2_request_list_lock); \ |
| 700 | spin_unlock(&op->lock); \ |
| 701 | wake_up_interruptible(&pvfs2_request_list_waitq); \ |
| 702 | } while (0) |
| 703 | |
| 704 | #define add_priority_op_to_request_list(op) \ |
| 705 | do { \ |
| 706 | spin_lock(&pvfs2_request_list_lock); \ |
| 707 | spin_lock(&op->lock); \ |
| 708 | set_op_state_waiting(op); \ |
| 709 | \ |
| 710 | list_add(&op->list, &pvfs2_request_list); \ |
| 711 | spin_unlock(&pvfs2_request_list_lock); \ |
| 712 | spin_unlock(&op->lock); \ |
| 713 | wake_up_interruptible(&pvfs2_request_list_waitq); \ |
| 714 | } while (0) |
| 715 | |
| 716 | #define remove_op_from_request_list(op) \ |
| 717 | do { \ |
| 718 | struct list_head *tmp = NULL; \ |
| 719 | struct list_head *tmp_safe = NULL; \ |
| 720 | struct pvfs2_kernel_op_s *tmp_op = NULL; \ |
| 721 | \ |
| 722 | spin_lock(&pvfs2_request_list_lock); \ |
| 723 | list_for_each_safe(tmp, tmp_safe, &pvfs2_request_list) { \ |
| 724 | tmp_op = list_entry(tmp, \ |
| 725 | struct pvfs2_kernel_op_s, \ |
| 726 | list); \ |
| 727 | if (tmp_op && (tmp_op == op)) { \ |
| 728 | list_del(&tmp_op->list); \ |
| 729 | break; \ |
| 730 | } \ |
| 731 | } \ |
| 732 | spin_unlock(&pvfs2_request_list_lock); \ |
| 733 | } while (0) |
| 734 | |
| 735 | #define PVFS2_OP_INTERRUPTIBLE 1 /* service_operation() is interruptible */ |
| 736 | #define PVFS2_OP_PRIORITY 2 /* service_operation() is high priority */ |
| 737 | #define PVFS2_OP_CANCELLATION 4 /* this is a cancellation */ |
| 738 | #define PVFS2_OP_NO_SEMAPHORE 8 /* don't acquire semaphore */ |
| 739 | #define PVFS2_OP_ASYNC 16 /* Queue it, but don't wait */ |
| 740 | |
| 741 | int service_operation(struct pvfs2_kernel_op_s *op, |
| 742 | const char *op_name, |
| 743 | int flags); |
| 744 | |
| 745 | /* |
| 746 | * handles two possible error cases, depending on context. |
| 747 | * |
| 748 | * by design, our vfs i/o errors need to be handled in one of two ways, |
| 749 | * depending on where the error occured. |
| 750 | * |
| 751 | * if the error happens in the waitqueue code because we either timed |
| 752 | * out or a signal was raised while waiting, we need to cancel the |
| 753 | * userspace i/o operation and free the op manually. this is done to |
| 754 | * avoid having the device start writing application data to our shared |
| 755 | * bufmap pages without us expecting it. |
| 756 | * |
| 757 | * FIXME: POSSIBLE OPTIMIZATION: |
| 758 | * However, if we timed out or if we got a signal AND our upcall was never |
| 759 | * picked off the queue (i.e. we were in OP_VFS_STATE_WAITING), then we don't |
| 760 | * need to send a cancellation upcall. The way we can handle this is |
| 761 | * set error_exit to 2 in such cases and 1 whenever cancellation has to be |
| 762 | * sent and have handle_error |
| 763 | * take care of this situation as well.. |
| 764 | * |
| 765 | * if a pvfs2 sysint level error occured and i/o has been completed, |
| 766 | * there is no need to cancel the operation, as the user has finished |
| 767 | * using the bufmap page and so there is no danger in this case. in |
| 768 | * this case, we wake up the device normally so that it may free the |
| 769 | * op, as normal. |
| 770 | * |
| 771 | * note the only reason this is a macro is because both read and write |
| 772 | * cases need the exact same handling code. |
| 773 | */ |
| 774 | #define handle_io_error() \ |
| 775 | do { \ |
| 776 | if (!op_state_serviced(new_op)) { \ |
| 777 | pvfs2_cancel_op_in_progress(new_op->tag); \ |
| 778 | op_release(new_op); \ |
| 779 | } else { \ |
| 780 | wake_up_daemon_for_return(new_op); \ |
| 781 | } \ |
| 782 | new_op = NULL; \ |
| 783 | pvfs_bufmap_put(bufmap, buffer_index); \ |
| 784 | buffer_index = -1; \ |
| 785 | } while (0) |
| 786 | |
| 787 | #define get_interruptible_flag(inode) \ |
| 788 | ((PVFS2_SB(inode->i_sb)->flags & PVFS2_OPT_INTR) ? \ |
| 789 | PVFS2_OP_INTERRUPTIBLE : 0) |
| 790 | |
| 791 | #define add_pvfs2_sb(sb) \ |
| 792 | do { \ |
| 793 | gossip_debug(GOSSIP_SUPER_DEBUG, \ |
| 794 | "Adding SB %p to pvfs2 superblocks\n", \ |
| 795 | PVFS2_SB(sb)); \ |
| 796 | spin_lock(&pvfs2_superblocks_lock); \ |
| 797 | list_add_tail(&PVFS2_SB(sb)->list, &pvfs2_superblocks); \ |
| 798 | spin_unlock(&pvfs2_superblocks_lock); \ |
| 799 | } while (0) |
| 800 | |
| 801 | #define remove_pvfs2_sb(sb) \ |
| 802 | do { \ |
| 803 | struct list_head *tmp = NULL; \ |
| 804 | struct list_head *tmp_safe = NULL; \ |
| 805 | struct pvfs2_sb_info_s *pvfs2_sb = NULL; \ |
| 806 | \ |
| 807 | spin_lock(&pvfs2_superblocks_lock); \ |
| 808 | list_for_each_safe(tmp, tmp_safe, &pvfs2_superblocks) { \ |
| 809 | pvfs2_sb = list_entry(tmp, \ |
| 810 | struct pvfs2_sb_info_s, \ |
| 811 | list); \ |
| 812 | if (pvfs2_sb && (pvfs2_sb->sb == sb)) { \ |
| 813 | gossip_debug(GOSSIP_SUPER_DEBUG, \ |
| 814 | "Removing SB %p from pvfs2 superblocks\n", \ |
| 815 | pvfs2_sb); \ |
| 816 | list_del(&pvfs2_sb->list); \ |
| 817 | break; \ |
| 818 | } \ |
| 819 | } \ |
| 820 | spin_unlock(&pvfs2_superblocks_lock); \ |
| 821 | } while (0) |
| 822 | |
| 823 | #define pvfs2_lock_inode(inode) spin_lock(&inode->i_lock) |
| 824 | #define pvfs2_unlock_inode(inode) spin_unlock(&inode->i_lock) |
| 825 | #define pvfs2_current_signal_lock current->sighand->siglock |
| 826 | #define pvfs2_current_sigaction current->sighand->action |
| 827 | |
| 828 | #define fill_default_sys_attrs(sys_attr, type, mode) \ |
| 829 | do { \ |
| 830 | sys_attr.owner = from_kuid(current_user_ns(), current_fsuid()); \ |
| 831 | sys_attr.group = from_kgid(current_user_ns(), current_fsgid()); \ |
| 832 | sys_attr.size = 0; \ |
| 833 | sys_attr.perms = PVFS_util_translate_mode(mode); \ |
| 834 | sys_attr.objtype = type; \ |
| 835 | sys_attr.mask = PVFS_ATTR_SYS_ALL_SETABLE; \ |
| 836 | } while (0) |
| 837 | |
| 838 | #define pvfs2_inode_lock(__i) mutex_lock(&(__i)->i_mutex) |
| 839 | |
| 840 | #define pvfs2_inode_unlock(__i) mutex_unlock(&(__i)->i_mutex) |
| 841 | |
| 842 | static inline void pvfs2_i_size_write(struct inode *inode, loff_t i_size) |
| 843 | { |
| 844 | #if BITS_PER_LONG == 32 && defined(CONFIG_SMP) |
| 845 | pvfs2_inode_lock(inode); |
| 846 | #endif |
| 847 | i_size_write(inode, i_size); |
| 848 | #if BITS_PER_LONG == 32 && defined(CONFIG_SMP) |
| 849 | pvfs2_inode_unlock(inode); |
| 850 | #endif |
| 851 | } |
| 852 | |
| 853 | static inline unsigned int diff(struct timeval *end, struct timeval *begin) |
| 854 | { |
| 855 | if (end->tv_usec < begin->tv_usec) { |
| 856 | end->tv_usec += 1000000; |
| 857 | end->tv_sec--; |
| 858 | } |
| 859 | end->tv_sec -= begin->tv_sec; |
| 860 | end->tv_usec -= begin->tv_usec; |
| 861 | return (end->tv_sec * 1000000) + end->tv_usec; |
| 862 | } |
| 863 | |
| 864 | #endif /* __PVFS2KERNEL_H */ |