Commit | Line | Data |
---|---|---|
f7ab093f MM |
1 | /* |
2 | * (C) 2001 Clemson University and The University of Chicago | |
3 | * | |
4 | * See COPYING in top-level directory. | |
5 | */ | |
6 | ||
7 | /* | |
8bb8aefd | 8 | * The ORANGEFS Linux kernel support allows ORANGEFS volumes to be mounted and |
f7ab093f MM |
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 | /* | |
8bb8aefd | 15 | * Declarations and macros for the ORANGEFS Linux kernel support. |
f7ab093f MM |
16 | */ |
17 | ||
8bb8aefd YL |
18 | #ifndef __ORANGEFSKERNEL_H |
19 | #define __ORANGEFSKERNEL_H | |
f7ab093f MM |
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 | ||
575e9461 | 56 | #include "orangefs-dev-proto.h" |
f7ab093f | 57 | |
8bb8aefd YL |
58 | #ifdef ORANGEFS_KERNEL_DEBUG |
59 | #define ORANGEFS_DEFAULT_OP_TIMEOUT_SECS 10 | |
f7ab093f | 60 | #else |
8bb8aefd | 61 | #define ORANGEFS_DEFAULT_OP_TIMEOUT_SECS 20 |
f7ab093f MM |
62 | #endif |
63 | ||
8bb8aefd | 64 | #define ORANGEFS_BUFMAP_WAIT_TIMEOUT_SECS 30 |
f7ab093f | 65 | |
8bb8aefd | 66 | #define ORANGEFS_DEFAULT_SLOT_TIMEOUT_SECS 900 /* 15 minutes */ |
f7ab093f | 67 | |
8bb8aefd | 68 | #define ORANGEFS_REQDEVICE_NAME "pvfs2-req" |
f7ab093f | 69 | |
8bb8aefd YL |
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 | |
f7ab093f | 77 | |
cf0c2771 | 78 | #define MAX_DEV_REQ_UPSIZE (2 * sizeof(__s32) + \ |
8bb8aefd | 79 | sizeof(__u64) + sizeof(struct orangefs_upcall_s)) |
cf0c2771 | 80 | #define MAX_DEV_REQ_DOWNSIZE (2 * sizeof(__s32) + \ |
8bb8aefd | 81 | sizeof(__u64) + sizeof(struct orangefs_downcall_s)) |
f7ab093f | 82 | |
f7ab093f MM |
83 | /* borrowed from irda.h */ |
84 | #ifndef MSECS_TO_JIFFIES | |
85 | #define MSECS_TO_JIFFIES(ms) (((ms)*HZ+999)/1000) | |
86 | #endif | |
87 | ||
f7ab093f | 88 | /* |
8bb8aefd | 89 | * valid orangefs kernel operation states |
f7ab093f MM |
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 | */ | |
8bb8aefd | 98 | enum orangefs_vfs_op_states { |
f7ab093f MM |
99 | OP_VFS_STATE_UNKNOWN = 0, |
100 | OP_VFS_STATE_WAITING = 1, | |
101 | OP_VFS_STATE_INPROGR = 2, | |
102 | OP_VFS_STATE_SERVICED = 4, | |
103 | OP_VFS_STATE_PURGED = 8, | |
104 | }; | |
105 | ||
f7ab093f MM |
106 | #define get_op(op) \ |
107 | do { \ | |
c817e266 | 108 | atomic_inc(&(op)->ref_count); \ |
f7ab093f MM |
109 | gossip_debug(GOSSIP_DEV_DEBUG, \ |
110 | "(get) Alloced OP (%p:%llu)\n", \ | |
111 | op, \ | |
112 | llu((op)->tag)); \ | |
113 | } while (0) | |
114 | ||
115 | #define put_op(op) \ | |
116 | do { \ | |
c817e266 | 117 | if (atomic_sub_and_test(1, &(op)->ref_count) == 1) { \ |
f7ab093f MM |
118 | gossip_debug(GOSSIP_DEV_DEBUG, \ |
119 | "(put) Releasing OP (%p:%llu)\n", \ | |
120 | op, \ | |
121 | llu((op)->tag)); \ | |
122 | op_release(op); \ | |
123 | } \ | |
124 | } while (0) | |
125 | ||
c817e266 | 126 | #define op_wait(op) (atomic_read(&(op)->ref_count) <= 2 ? 0 : 1) |
f7ab093f MM |
127 | |
128 | /* | |
129 | * Defines for controlling whether I/O upcalls are for async or sync operations | |
130 | */ | |
8bb8aefd YL |
131 | enum ORANGEFS_async_io_type { |
132 | ORANGEFS_VFS_SYNC_IO = 0, | |
133 | ORANGEFS_VFS_ASYNC_IO = 1, | |
f7ab093f MM |
134 | }; |
135 | ||
136 | /* | |
137 | * An array of client_debug_mask will be built to hold debug keyword/mask | |
138 | * values fetched from userspace. | |
139 | */ | |
140 | struct client_debug_mask { | |
141 | char *keyword; | |
142 | __u64 mask1; | |
143 | __u64 mask2; | |
144 | }; | |
145 | ||
146 | /* | |
8bb8aefd | 147 | * orangefs kernel memory related flags |
f7ab093f MM |
148 | */ |
149 | ||
8bb8aefd YL |
150 | #if ((defined ORANGEFS_KERNEL_DEBUG) && (defined CONFIG_DEBUG_SLAB)) |
151 | #define ORANGEFS_CACHE_CREATE_FLAGS SLAB_RED_ZONE | |
f7ab093f | 152 | #else |
8bb8aefd YL |
153 | #define ORANGEFS_CACHE_CREATE_FLAGS 0 |
154 | #endif /* ((defined ORANGEFS_KERNEL_DEBUG) && (defined CONFIG_DEBUG_SLAB)) */ | |
f7ab093f | 155 | |
8bb8aefd YL |
156 | #define ORANGEFS_CACHE_ALLOC_FLAGS (GFP_KERNEL) |
157 | #define ORANGEFS_GFP_FLAGS (GFP_KERNEL) | |
158 | #define ORANGEFS_BUFMAP_GFP_FLAGS (GFP_KERNEL) | |
f7ab093f | 159 | |
8bb8aefd YL |
160 | /* orangefs xattr and acl related defines */ |
161 | #define ORANGEFS_XATTR_INDEX_POSIX_ACL_ACCESS 1 | |
162 | #define ORANGEFS_XATTR_INDEX_POSIX_ACL_DEFAULT 2 | |
163 | #define ORANGEFS_XATTR_INDEX_TRUSTED 3 | |
164 | #define ORANGEFS_XATTR_INDEX_DEFAULT 4 | |
f7ab093f | 165 | |
fef8b67c MM |
166 | #define ORANGEFS_XATTR_NAME_ACL_ACCESS XATTR_NAME_POSIX_ACL_ACCESS |
167 | #define ORANGEFS_XATTR_NAME_ACL_DEFAULT XATTR_NAME_POSIX_ACL_DEFAULT | |
8bb8aefd YL |
168 | #define ORANGEFS_XATTR_NAME_TRUSTED_PREFIX "trusted." |
169 | #define ORANGEFS_XATTR_NAME_DEFAULT_PREFIX "" | |
f7ab093f | 170 | |
8bb8aefd | 171 | /* these functions are defined in orangefs-utils.c */ |
f7ab093f MM |
172 | int orangefs_prepare_cdm_array(char *debug_array_string); |
173 | int orangefs_prepare_debugfs_help_string(int); | |
174 | ||
8bb8aefd YL |
175 | /* defined in orangefs-debugfs.c */ |
176 | int orangefs_client_debug_init(void); | |
f7ab093f MM |
177 | |
178 | void debug_string_to_mask(char *, void *, int); | |
179 | void do_c_mask(int, char *, struct client_debug_mask **); | |
180 | void do_k_mask(int, char *, __u64 **); | |
181 | ||
182 | void debug_mask_to_string(void *, int); | |
183 | void do_k_string(void *, int); | |
184 | void do_c_string(void *, int); | |
185 | int check_amalgam_keyword(void *, int); | |
186 | int keyword_is_amalgam(char *); | |
187 | ||
8bb8aefd YL |
188 | /*these variables are defined in orangefs-mod.c */ |
189 | extern char kernel_debug_string[ORANGEFS_MAX_DEBUG_STRING_LEN]; | |
190 | extern char client_debug_string[ORANGEFS_MAX_DEBUG_STRING_LEN]; | |
191 | extern char client_debug_array_string[ORANGEFS_MAX_DEBUG_STRING_LEN]; | |
f7ab093f MM |
192 | extern unsigned int kernel_mask_set_mod_init; |
193 | ||
8bb8aefd YL |
194 | extern int orangefs_init_acl(struct inode *inode, struct inode *dir); |
195 | extern const struct xattr_handler *orangefs_xattr_handlers[]; | |
f7ab093f | 196 | |
8bb8aefd YL |
197 | extern struct posix_acl *orangefs_get_acl(struct inode *inode, int type); |
198 | extern int orangefs_set_acl(struct inode *inode, struct posix_acl *acl, int type); | |
f7ab093f | 199 | |
f7ab093f MM |
200 | /* |
201 | * Redefine xtvec structure so that we could move helper functions out of | |
202 | * the define | |
203 | */ | |
204 | struct xtvec { | |
205 | __kernel_off_t xtv_off; /* must be off_t */ | |
206 | __kernel_size_t xtv_len; /* must be size_t */ | |
207 | }; | |
208 | ||
209 | /* | |
8bb8aefd | 210 | * orangefs data structures |
f7ab093f | 211 | */ |
8bb8aefd YL |
212 | struct orangefs_kernel_op_s { |
213 | enum orangefs_vfs_op_states op_state; | |
f7ab093f MM |
214 | __u64 tag; |
215 | ||
216 | /* | |
217 | * Set uses_shared_memory to 1 if this operation uses shared memory. | |
218 | * If true, then a retry on the op must also get a new shared memory | |
219 | * buffer and re-populate it. | |
220 | */ | |
221 | int uses_shared_memory; | |
222 | ||
8bb8aefd YL |
223 | struct orangefs_upcall_s upcall; |
224 | struct orangefs_downcall_s downcall; | |
f7ab093f MM |
225 | |
226 | wait_queue_head_t waitq; | |
227 | spinlock_t lock; | |
228 | ||
229 | int io_completed; | |
230 | wait_queue_head_t io_completion_waitq; | |
231 | ||
c817e266 MM |
232 | atomic_t ref_count; |
233 | ||
f7ab093f MM |
234 | /* VFS aio fields */ |
235 | ||
8bb8aefd | 236 | /* used by the async I/O code to stash the orangefs_kiocb_s structure */ |
f7ab093f MM |
237 | void *priv; |
238 | ||
f7ab093f MM |
239 | int attempts; |
240 | ||
241 | struct list_head list; | |
242 | }; | |
243 | ||
60831949 AV |
244 | #define set_op_state_waiting(op) ((op)->op_state = OP_VFS_STATE_WAITING) |
245 | #define set_op_state_inprogress(op) ((op)->op_state = OP_VFS_STATE_INPROGR) | |
246 | static inline void set_op_state_serviced(struct orangefs_kernel_op_s *op) | |
247 | { | |
248 | op->op_state = OP_VFS_STATE_SERVICED; | |
249 | wake_up_interruptible(&op->waitq); | |
250 | } | |
251 | static inline void set_op_state_purged(struct orangefs_kernel_op_s *op) | |
252 | { | |
253 | op->op_state |= OP_VFS_STATE_PURGED; | |
254 | wake_up_interruptible(&op->waitq); | |
255 | } | |
256 | ||
257 | #define op_state_waiting(op) ((op)->op_state & OP_VFS_STATE_WAITING) | |
258 | #define op_state_in_progress(op) ((op)->op_state & OP_VFS_STATE_INPROGR) | |
259 | #define op_state_serviced(op) ((op)->op_state & OP_VFS_STATE_SERVICED) | |
260 | #define op_state_purged(op) ((op)->op_state & OP_VFS_STATE_PURGED) | |
261 | ||
8bb8aefd YL |
262 | /* per inode private orangefs info */ |
263 | struct orangefs_inode_s { | |
264 | struct orangefs_object_kref refn; | |
265 | char link_target[ORANGEFS_NAME_MAX]; | |
f7ab093f MM |
266 | __s64 blksize; |
267 | /* | |
268 | * Reading/Writing Extended attributes need to acquire the appropriate | |
8bb8aefd | 269 | * reader/writer semaphore on the orangefs_inode_s structure. |
f7ab093f MM |
270 | */ |
271 | struct rw_semaphore xattr_sem; | |
272 | ||
273 | struct inode vfs_inode; | |
274 | sector_t last_failed_block_index_read; | |
275 | ||
276 | /* | |
277 | * State of in-memory attributes not yet flushed to disk associated | |
278 | * with this object | |
279 | */ | |
280 | unsigned long pinode_flags; | |
f7ab093f MM |
281 | }; |
282 | ||
283 | #define P_ATIME_FLAG 0 | |
284 | #define P_MTIME_FLAG 1 | |
285 | #define P_CTIME_FLAG 2 | |
286 | #define P_MODE_FLAG 3 | |
287 | ||
288 | #define ClearAtimeFlag(pinode) clear_bit(P_ATIME_FLAG, &(pinode)->pinode_flags) | |
289 | #define SetAtimeFlag(pinode) set_bit(P_ATIME_FLAG, &(pinode)->pinode_flags) | |
290 | #define AtimeFlag(pinode) test_bit(P_ATIME_FLAG, &(pinode)->pinode_flags) | |
291 | ||
292 | #define ClearMtimeFlag(pinode) clear_bit(P_MTIME_FLAG, &(pinode)->pinode_flags) | |
293 | #define SetMtimeFlag(pinode) set_bit(P_MTIME_FLAG, &(pinode)->pinode_flags) | |
294 | #define MtimeFlag(pinode) test_bit(P_MTIME_FLAG, &(pinode)->pinode_flags) | |
295 | ||
296 | #define ClearCtimeFlag(pinode) clear_bit(P_CTIME_FLAG, &(pinode)->pinode_flags) | |
297 | #define SetCtimeFlag(pinode) set_bit(P_CTIME_FLAG, &(pinode)->pinode_flags) | |
298 | #define CtimeFlag(pinode) test_bit(P_CTIME_FLAG, &(pinode)->pinode_flags) | |
299 | ||
300 | #define ClearModeFlag(pinode) clear_bit(P_MODE_FLAG, &(pinode)->pinode_flags) | |
301 | #define SetModeFlag(pinode) set_bit(P_MODE_FLAG, &(pinode)->pinode_flags) | |
302 | #define ModeFlag(pinode) test_bit(P_MODE_FLAG, &(pinode)->pinode_flags) | |
303 | ||
8bb8aefd YL |
304 | /* per superblock private orangefs info */ |
305 | struct orangefs_sb_info_s { | |
306 | struct orangefs_khandle root_khandle; | |
f7ab093f MM |
307 | __s32 fs_id; |
308 | int id; | |
309 | int flags; | |
8bb8aefd YL |
310 | #define ORANGEFS_OPT_INTR 0x01 |
311 | #define ORANGEFS_OPT_LOCAL_LOCK 0x02 | |
312 | char devname[ORANGEFS_MAX_SERVER_ADDR_LEN]; | |
f7ab093f MM |
313 | struct super_block *sb; |
314 | int mount_pending; | |
315 | struct list_head list; | |
316 | }; | |
317 | ||
f7ab093f MM |
318 | /* |
319 | * structure that holds the state of any async I/O operation issued | |
320 | * through the VFS. Needed especially to handle cancellation requests | |
321 | * or even completion notification so that the VFS client-side daemon | |
322 | * can free up its vfs_request slots. | |
323 | */ | |
8bb8aefd | 324 | struct orangefs_kiocb_s { |
f7ab093f MM |
325 | /* the pointer to the task that initiated the AIO */ |
326 | struct task_struct *tsk; | |
327 | ||
328 | /* pointer to the kiocb that kicked this operation */ | |
329 | struct kiocb *kiocb; | |
330 | ||
331 | /* buffer index that was used for the I/O */ | |
8bb8aefd | 332 | struct orangefs_bufmap *bufmap; |
f7ab093f MM |
333 | int buffer_index; |
334 | ||
8bb8aefd YL |
335 | /* orangefs kernel operation type */ |
336 | struct orangefs_kernel_op_s *op; | |
f7ab093f MM |
337 | |
338 | /* The user space buffers from/to which I/O is being staged */ | |
339 | struct iovec *iov; | |
340 | ||
341 | /* number of elements in the iovector */ | |
342 | unsigned long nr_segs; | |
343 | ||
344 | /* set to indicate the type of the operation */ | |
345 | int rw; | |
346 | ||
347 | /* file offset */ | |
348 | loff_t offset; | |
349 | ||
350 | /* and the count in bytes */ | |
351 | size_t bytes_to_be_copied; | |
352 | ||
353 | ssize_t bytes_copied; | |
354 | int needs_cleanup; | |
355 | }; | |
356 | ||
8bb8aefd | 357 | struct orangefs_stats { |
f7ab093f MM |
358 | unsigned long cache_hits; |
359 | unsigned long cache_misses; | |
360 | unsigned long reads; | |
361 | unsigned long writes; | |
362 | }; | |
363 | ||
8bb8aefd | 364 | extern struct orangefs_stats g_orangefs_stats; |
f7ab093f MM |
365 | |
366 | /* | |
54804949 MM |
367 | * NOTE: See Documentation/filesystems/porting for information |
368 | * on implementing FOO_I and properly accessing fs private data | |
369 | */ | |
8bb8aefd | 370 | static inline struct orangefs_inode_s *ORANGEFS_I(struct inode *inode) |
f7ab093f | 371 | { |
8bb8aefd | 372 | return container_of(inode, struct orangefs_inode_s, vfs_inode); |
f7ab093f MM |
373 | } |
374 | ||
8bb8aefd | 375 | static inline struct orangefs_sb_info_s *ORANGEFS_SB(struct super_block *sb) |
f7ab093f | 376 | { |
8bb8aefd | 377 | return (struct orangefs_sb_info_s *) sb->s_fs_info; |
f7ab093f MM |
378 | } |
379 | ||
380 | /* ino_t descends from "unsigned long", 8 bytes, 64 bits. */ | |
8bb8aefd | 381 | static inline ino_t orangefs_khandle_to_ino(struct orangefs_khandle *khandle) |
f7ab093f MM |
382 | { |
383 | union { | |
384 | unsigned char u[8]; | |
385 | __u64 ino; | |
386 | } ihandle; | |
387 | ||
388 | ihandle.u[0] = khandle->u[0] ^ khandle->u[4]; | |
389 | ihandle.u[1] = khandle->u[1] ^ khandle->u[5]; | |
390 | ihandle.u[2] = khandle->u[2] ^ khandle->u[6]; | |
391 | ihandle.u[3] = khandle->u[3] ^ khandle->u[7]; | |
392 | ihandle.u[4] = khandle->u[12] ^ khandle->u[8]; | |
393 | ihandle.u[5] = khandle->u[13] ^ khandle->u[9]; | |
394 | ihandle.u[6] = khandle->u[14] ^ khandle->u[10]; | |
395 | ihandle.u[7] = khandle->u[15] ^ khandle->u[11]; | |
396 | ||
397 | return ihandle.ino; | |
398 | } | |
399 | ||
8bb8aefd | 400 | static inline struct orangefs_khandle *get_khandle_from_ino(struct inode *inode) |
f7ab093f | 401 | { |
8bb8aefd | 402 | return &(ORANGEFS_I(inode)->refn.khandle); |
f7ab093f MM |
403 | } |
404 | ||
405 | static inline __s32 get_fsid_from_ino(struct inode *inode) | |
406 | { | |
8bb8aefd | 407 | return ORANGEFS_I(inode)->refn.fs_id; |
f7ab093f MM |
408 | } |
409 | ||
410 | static inline ino_t get_ino_from_khandle(struct inode *inode) | |
411 | { | |
8bb8aefd | 412 | struct orangefs_khandle *khandle; |
f7ab093f MM |
413 | ino_t ino; |
414 | ||
415 | khandle = get_khandle_from_ino(inode); | |
8bb8aefd | 416 | ino = orangefs_khandle_to_ino(khandle); |
f7ab093f MM |
417 | return ino; |
418 | } | |
419 | ||
420 | static inline ino_t get_parent_ino_from_dentry(struct dentry *dentry) | |
421 | { | |
422 | return get_ino_from_khandle(dentry->d_parent->d_inode); | |
423 | } | |
424 | ||
425 | static inline int is_root_handle(struct inode *inode) | |
426 | { | |
427 | gossip_debug(GOSSIP_DCACHE_DEBUG, | |
428 | "%s: root handle: %pU, this handle: %pU:\n", | |
429 | __func__, | |
8bb8aefd | 430 | &ORANGEFS_SB(inode->i_sb)->root_khandle, |
f7ab093f MM |
431 | get_khandle_from_ino(inode)); |
432 | ||
8bb8aefd | 433 | if (ORANGEFS_khandle_cmp(&(ORANGEFS_SB(inode->i_sb)->root_khandle), |
f7ab093f MM |
434 | get_khandle_from_ino(inode))) |
435 | return 0; | |
436 | else | |
437 | return 1; | |
438 | } | |
439 | ||
8bb8aefd | 440 | static inline int match_handle(struct orangefs_khandle resp_handle, |
f7ab093f MM |
441 | struct inode *inode) |
442 | { | |
443 | gossip_debug(GOSSIP_DCACHE_DEBUG, | |
444 | "%s: one handle: %pU, another handle:%pU:\n", | |
445 | __func__, | |
446 | &resp_handle, | |
447 | get_khandle_from_ino(inode)); | |
448 | ||
8bb8aefd | 449 | if (ORANGEFS_khandle_cmp(&resp_handle, get_khandle_from_ino(inode))) |
f7ab093f MM |
450 | return 0; |
451 | else | |
452 | return 1; | |
453 | } | |
454 | ||
455 | /* | |
8bb8aefd | 456 | * defined in orangefs-cache.c |
f7ab093f MM |
457 | */ |
458 | int op_cache_initialize(void); | |
459 | int op_cache_finalize(void); | |
8bb8aefd YL |
460 | struct orangefs_kernel_op_s *op_alloc(__s32 type); |
461 | char *get_opname_string(struct orangefs_kernel_op_s *new_op); | |
462 | void op_release(struct orangefs_kernel_op_s *op); | |
f7ab093f MM |
463 | |
464 | int dev_req_cache_initialize(void); | |
465 | int dev_req_cache_finalize(void); | |
466 | void *dev_req_alloc(void); | |
467 | void dev_req_release(void *); | |
468 | ||
8bb8aefd YL |
469 | int orangefs_inode_cache_initialize(void); |
470 | int orangefs_inode_cache_finalize(void); | |
f7ab093f MM |
471 | |
472 | int kiocb_cache_initialize(void); | |
473 | int kiocb_cache_finalize(void); | |
8bb8aefd YL |
474 | struct orangefs_kiocb_s *kiocb_alloc(void); |
475 | void kiocb_release(struct orangefs_kiocb_s *ptr); | |
f7ab093f MM |
476 | |
477 | /* | |
8bb8aefd | 478 | * defined in orangefs-mod.c |
f7ab093f MM |
479 | */ |
480 | void purge_inprogress_ops(void); | |
481 | ||
482 | /* | |
483 | * defined in waitqueue.c | |
484 | */ | |
f7ab093f MM |
485 | void purge_waiting_ops(void); |
486 | ||
487 | /* | |
488 | * defined in super.c | |
489 | */ | |
8bb8aefd | 490 | struct dentry *orangefs_mount(struct file_system_type *fst, |
f7ab093f MM |
491 | int flags, |
492 | const char *devname, | |
493 | void *data); | |
494 | ||
8bb8aefd YL |
495 | void orangefs_kill_sb(struct super_block *sb); |
496 | int orangefs_remount(struct super_block *sb); | |
f7ab093f MM |
497 | |
498 | int fsid_key_table_initialize(void); | |
499 | void fsid_key_table_finalize(void); | |
500 | ||
501 | /* | |
502 | * defined in inode.c | |
503 | */ | |
8bb8aefd YL |
504 | __u32 convert_to_orangefs_mask(unsigned long lite_mask); |
505 | struct inode *orangefs_new_inode(struct super_block *sb, | |
f7ab093f MM |
506 | struct inode *dir, |
507 | int mode, | |
508 | dev_t dev, | |
8bb8aefd | 509 | struct orangefs_object_kref *ref); |
f7ab093f | 510 | |
8bb8aefd | 511 | int orangefs_setattr(struct dentry *dentry, struct iattr *iattr); |
f7ab093f | 512 | |
8bb8aefd | 513 | int orangefs_getattr(struct vfsmount *mnt, |
f7ab093f MM |
514 | struct dentry *dentry, |
515 | struct kstat *kstat); | |
516 | ||
517 | /* | |
518 | * defined in xattr.c | |
519 | */ | |
8bb8aefd | 520 | int orangefs_setxattr(struct dentry *dentry, |
f7ab093f MM |
521 | const char *name, |
522 | const void *value, | |
523 | size_t size, | |
524 | int flags); | |
525 | ||
8bb8aefd | 526 | ssize_t orangefs_getxattr(struct dentry *dentry, |
f7ab093f MM |
527 | const char *name, |
528 | void *buffer, | |
529 | size_t size); | |
530 | ||
8bb8aefd | 531 | ssize_t orangefs_listxattr(struct dentry *dentry, char *buffer, size_t size); |
f7ab093f MM |
532 | |
533 | /* | |
534 | * defined in namei.c | |
535 | */ | |
8bb8aefd YL |
536 | struct inode *orangefs_iget(struct super_block *sb, |
537 | struct orangefs_object_kref *ref); | |
f7ab093f | 538 | |
8bb8aefd YL |
539 | ssize_t orangefs_inode_read(struct inode *inode, |
540 | struct iov_iter *iter, | |
541 | loff_t *offset, | |
542 | loff_t readahead_size); | |
f7ab093f MM |
543 | |
544 | /* | |
8bb8aefd | 545 | * defined in devorangefs-req.c |
f7ab093f | 546 | */ |
8bb8aefd YL |
547 | int orangefs_dev_init(void); |
548 | void orangefs_dev_cleanup(void); | |
f7ab093f MM |
549 | int is_daemon_in_service(void); |
550 | int fs_mount_pending(__s32 fsid); | |
551 | ||
552 | /* | |
8bb8aefd | 553 | * defined in orangefs-utils.c |
f7ab093f | 554 | */ |
8bb8aefd | 555 | __s32 fsid_of_op(struct orangefs_kernel_op_s *op); |
f7ab093f | 556 | |
8bb8aefd | 557 | int orangefs_flush_inode(struct inode *inode); |
f7ab093f | 558 | |
8bb8aefd | 559 | ssize_t orangefs_inode_getxattr(struct inode *inode, |
f7ab093f MM |
560 | const char *prefix, |
561 | const char *name, | |
562 | void *buffer, | |
563 | size_t size); | |
564 | ||
8bb8aefd | 565 | int orangefs_inode_setxattr(struct inode *inode, |
f7ab093f MM |
566 | const char *prefix, |
567 | const char *name, | |
568 | const void *value, | |
569 | size_t size, | |
570 | int flags); | |
571 | ||
8bb8aefd | 572 | int orangefs_inode_getattr(struct inode *inode, __u32 mask); |
f7ab093f | 573 | |
8bb8aefd | 574 | int orangefs_inode_setattr(struct inode *inode, struct iattr *iattr); |
f7ab093f | 575 | |
8bb8aefd | 576 | void orangefs_op_initialize(struct orangefs_kernel_op_s *op); |
f7ab093f | 577 | |
8bb8aefd | 578 | void orangefs_make_bad_inode(struct inode *inode); |
f7ab093f | 579 | |
c146c0b8 | 580 | void orangefs_block_signals(sigset_t *); |
f7ab093f | 581 | |
c146c0b8 | 582 | void orangefs_set_signals(sigset_t *); |
f7ab093f | 583 | |
8bb8aefd | 584 | int orangefs_unmount_sb(struct super_block *sb); |
f7ab093f | 585 | |
8bb8aefd | 586 | int orangefs_cancel_op_in_progress(__u64 tag); |
f7ab093f | 587 | |
8bb8aefd | 588 | static inline __u64 orangefs_convert_time_field(const struct timespec *ts) |
5714156b AV |
589 | { |
590 | return (__u64)ts->tv_sec; | |
591 | } | |
f7ab093f | 592 | |
8bb8aefd | 593 | int orangefs_normalize_to_errno(__s32 error_code); |
f7ab093f MM |
594 | |
595 | extern struct mutex devreq_mutex; | |
596 | extern struct mutex request_mutex; | |
597 | extern int debug; | |
598 | extern int op_timeout_secs; | |
599 | extern int slot_timeout_secs; | |
8bb8aefd YL |
600 | extern struct list_head orangefs_superblocks; |
601 | extern spinlock_t orangefs_superblocks_lock; | |
602 | extern struct list_head orangefs_request_list; | |
603 | extern spinlock_t orangefs_request_list_lock; | |
604 | extern wait_queue_head_t orangefs_request_list_waitq; | |
f7ab093f MM |
605 | extern struct list_head *htable_ops_in_progress; |
606 | extern spinlock_t htable_ops_in_progress_lock; | |
607 | extern int hash_table_size; | |
608 | ||
8bb8aefd YL |
609 | extern const struct address_space_operations orangefs_address_operations; |
610 | extern struct backing_dev_info orangefs_backing_dev_info; | |
611 | extern struct inode_operations orangefs_file_inode_operations; | |
612 | extern const struct file_operations orangefs_file_operations; | |
613 | extern struct inode_operations orangefs_symlink_inode_operations; | |
614 | extern struct inode_operations orangefs_dir_inode_operations; | |
615 | extern const struct file_operations orangefs_dir_operations; | |
616 | extern const struct dentry_operations orangefs_dentry_operations; | |
617 | extern const struct file_operations orangefs_devreq_file_operations; | |
f7ab093f | 618 | |
8bb8aefd | 619 | extern wait_queue_head_t orangefs_bufmap_init_waitq; |
f7ab093f MM |
620 | |
621 | /* | |
622 | * misc convenience macros | |
623 | */ | |
f7ab093f | 624 | |
8bb8aefd YL |
625 | #define ORANGEFS_OP_INTERRUPTIBLE 1 /* service_operation() is interruptible */ |
626 | #define ORANGEFS_OP_PRIORITY 2 /* service_operation() is high priority */ | |
627 | #define ORANGEFS_OP_CANCELLATION 4 /* this is a cancellation */ | |
628 | #define ORANGEFS_OP_NO_SEMAPHORE 8 /* don't acquire semaphore */ | |
629 | #define ORANGEFS_OP_ASYNC 16 /* Queue it, but don't wait */ | |
f7ab093f | 630 | |
8bb8aefd | 631 | int service_operation(struct orangefs_kernel_op_s *op, |
f7ab093f MM |
632 | const char *op_name, |
633 | int flags); | |
634 | ||
635 | /* | |
636 | * handles two possible error cases, depending on context. | |
637 | * | |
638 | * by design, our vfs i/o errors need to be handled in one of two ways, | |
639 | * depending on where the error occured. | |
640 | * | |
641 | * if the error happens in the waitqueue code because we either timed | |
642 | * out or a signal was raised while waiting, we need to cancel the | |
643 | * userspace i/o operation and free the op manually. this is done to | |
644 | * avoid having the device start writing application data to our shared | |
645 | * bufmap pages without us expecting it. | |
646 | * | |
647 | * FIXME: POSSIBLE OPTIMIZATION: | |
648 | * However, if we timed out or if we got a signal AND our upcall was never | |
649 | * picked off the queue (i.e. we were in OP_VFS_STATE_WAITING), then we don't | |
650 | * need to send a cancellation upcall. The way we can handle this is | |
651 | * set error_exit to 2 in such cases and 1 whenever cancellation has to be | |
652 | * sent and have handle_error | |
653 | * take care of this situation as well.. | |
654 | * | |
8bb8aefd | 655 | * if a orangefs sysint level error occured and i/o has been completed, |
f7ab093f MM |
656 | * there is no need to cancel the operation, as the user has finished |
657 | * using the bufmap page and so there is no danger in this case. in | |
658 | * this case, we wake up the device normally so that it may free the | |
659 | * op, as normal. | |
660 | * | |
661 | * note the only reason this is a macro is because both read and write | |
662 | * cases need the exact same handling code. | |
663 | */ | |
664 | #define handle_io_error() \ | |
665 | do { \ | |
666 | if (!op_state_serviced(new_op)) { \ | |
8bb8aefd | 667 | orangefs_cancel_op_in_progress(new_op->tag); \ |
f7ab093f MM |
668 | op_release(new_op); \ |
669 | } else { \ | |
670 | wake_up_daemon_for_return(new_op); \ | |
671 | } \ | |
672 | new_op = NULL; \ | |
8bb8aefd | 673 | orangefs_bufmap_put(bufmap, buffer_index); \ |
f7ab093f MM |
674 | buffer_index = -1; \ |
675 | } while (0) | |
676 | ||
677 | #define get_interruptible_flag(inode) \ | |
8bb8aefd YL |
678 | ((ORANGEFS_SB(inode->i_sb)->flags & ORANGEFS_OPT_INTR) ? \ |
679 | ORANGEFS_OP_INTERRUPTIBLE : 0) | |
f7ab093f | 680 | |
8bb8aefd | 681 | #define add_orangefs_sb(sb) \ |
f7ab093f MM |
682 | do { \ |
683 | gossip_debug(GOSSIP_SUPER_DEBUG, \ | |
8bb8aefd YL |
684 | "Adding SB %p to orangefs superblocks\n", \ |
685 | ORANGEFS_SB(sb)); \ | |
686 | spin_lock(&orangefs_superblocks_lock); \ | |
687 | list_add_tail(&ORANGEFS_SB(sb)->list, &orangefs_superblocks); \ | |
688 | spin_unlock(&orangefs_superblocks_lock); \ | |
f7ab093f MM |
689 | } while (0) |
690 | ||
8bb8aefd | 691 | #define remove_orangefs_sb(sb) \ |
f7ab093f MM |
692 | do { \ |
693 | struct list_head *tmp = NULL; \ | |
694 | struct list_head *tmp_safe = NULL; \ | |
8bb8aefd | 695 | struct orangefs_sb_info_s *orangefs_sb = NULL; \ |
f7ab093f | 696 | \ |
8bb8aefd YL |
697 | spin_lock(&orangefs_superblocks_lock); \ |
698 | list_for_each_safe(tmp, tmp_safe, &orangefs_superblocks) { \ | |
699 | orangefs_sb = list_entry(tmp, \ | |
700 | struct orangefs_sb_info_s, \ | |
f7ab093f | 701 | list); \ |
8bb8aefd | 702 | if (orangefs_sb && (orangefs_sb->sb == sb)) { \ |
f7ab093f | 703 | gossip_debug(GOSSIP_SUPER_DEBUG, \ |
8bb8aefd YL |
704 | "Removing SB %p from orangefs superblocks\n", \ |
705 | orangefs_sb); \ | |
706 | list_del(&orangefs_sb->list); \ | |
f7ab093f MM |
707 | break; \ |
708 | } \ | |
709 | } \ | |
8bb8aefd | 710 | spin_unlock(&orangefs_superblocks_lock); \ |
f7ab093f MM |
711 | } while (0) |
712 | ||
8bb8aefd YL |
713 | #define orangefs_lock_inode(inode) spin_lock(&inode->i_lock) |
714 | #define orangefs_unlock_inode(inode) spin_unlock(&inode->i_lock) | |
f7ab093f MM |
715 | |
716 | #define fill_default_sys_attrs(sys_attr, type, mode) \ | |
717 | do { \ | |
718 | sys_attr.owner = from_kuid(current_user_ns(), current_fsuid()); \ | |
719 | sys_attr.group = from_kgid(current_user_ns(), current_fsgid()); \ | |
720 | sys_attr.size = 0; \ | |
8bb8aefd | 721 | sys_attr.perms = ORANGEFS_util_translate_mode(mode); \ |
f7ab093f | 722 | sys_attr.objtype = type; \ |
8bb8aefd | 723 | sys_attr.mask = ORANGEFS_ATTR_SYS_ALL_SETABLE; \ |
f7ab093f MM |
724 | } while (0) |
725 | ||
8bb8aefd | 726 | #define orangefs_inode_lock(__i) mutex_lock(&(__i)->i_mutex) |
f7ab093f | 727 | |
8bb8aefd | 728 | #define orangefs_inode_unlock(__i) mutex_unlock(&(__i)->i_mutex) |
f7ab093f | 729 | |
8bb8aefd | 730 | static inline void orangefs_i_size_write(struct inode *inode, loff_t i_size) |
f7ab093f MM |
731 | { |
732 | #if BITS_PER_LONG == 32 && defined(CONFIG_SMP) | |
eb57bcc2 | 733 | orangefs_inode_lock(inode); |
f7ab093f MM |
734 | #endif |
735 | i_size_write(inode, i_size); | |
736 | #if BITS_PER_LONG == 32 && defined(CONFIG_SMP) | |
8bb8aefd | 737 | orangefs_inode_unlock(inode); |
f7ab093f MM |
738 | #endif |
739 | } | |
740 | ||
741 | static inline unsigned int diff(struct timeval *end, struct timeval *begin) | |
742 | { | |
743 | if (end->tv_usec < begin->tv_usec) { | |
744 | end->tv_usec += 1000000; | |
745 | end->tv_sec--; | |
746 | } | |
747 | end->tv_sec -= begin->tv_sec; | |
748 | end->tv_usec -= begin->tv_usec; | |
749 | return (end->tv_sec * 1000000) + end->tv_usec; | |
750 | } | |
751 | ||
8bb8aefd | 752 | #endif /* __ORANGEFSKERNEL_H */ |