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 | ||
56 | #include "pvfs2-dev-proto.h" | |
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 MM |
77 | |
78 | #define MAX_DEV_REQ_UPSIZE (2*sizeof(__s32) + \ | |
8bb8aefd | 79 | sizeof(__u64) + sizeof(struct orangefs_upcall_s)) |
f7ab093f | 80 | #define MAX_DEV_REQ_DOWNSIZE (2*sizeof(__s32) + \ |
8bb8aefd | 81 | sizeof(__u64) + sizeof(struct orangefs_downcall_s)) |
f7ab093f MM |
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 | /* | |
8bb8aefd | 107 | * valid orangefs kernel operation states |
f7ab093f MM |
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 | */ | |
8bb8aefd | 116 | enum orangefs_vfs_op_states { |
f7ab093f MM |
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 | */ | |
8bb8aefd YL |
159 | enum ORANGEFS_async_io_type { |
160 | ORANGEFS_VFS_SYNC_IO = 0, | |
161 | ORANGEFS_VFS_ASYNC_IO = 1, | |
f7ab093f MM |
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 | /* | |
8bb8aefd | 175 | * orangefs kernel memory related flags |
f7ab093f MM |
176 | */ |
177 | ||
8bb8aefd YL |
178 | #if ((defined ORANGEFS_KERNEL_DEBUG) && (defined CONFIG_DEBUG_SLAB)) |
179 | #define ORANGEFS_CACHE_CREATE_FLAGS SLAB_RED_ZONE | |
f7ab093f | 180 | #else |
8bb8aefd YL |
181 | #define ORANGEFS_CACHE_CREATE_FLAGS 0 |
182 | #endif /* ((defined ORANGEFS_KERNEL_DEBUG) && (defined CONFIG_DEBUG_SLAB)) */ | |
f7ab093f | 183 | |
8bb8aefd YL |
184 | #define ORANGEFS_CACHE_ALLOC_FLAGS (GFP_KERNEL) |
185 | #define ORANGEFS_GFP_FLAGS (GFP_KERNEL) | |
186 | #define ORANGEFS_BUFMAP_GFP_FLAGS (GFP_KERNEL) | |
f7ab093f | 187 | |
8bb8aefd YL |
188 | /* orangefs xattr and acl related defines */ |
189 | #define ORANGEFS_XATTR_INDEX_POSIX_ACL_ACCESS 1 | |
190 | #define ORANGEFS_XATTR_INDEX_POSIX_ACL_DEFAULT 2 | |
191 | #define ORANGEFS_XATTR_INDEX_TRUSTED 3 | |
192 | #define ORANGEFS_XATTR_INDEX_DEFAULT 4 | |
f7ab093f MM |
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 | ||
8bb8aefd YL |
203 | #define ORANGEFS_XATTR_NAME_ACL_ACCESS POSIX_ACL_XATTR_ACCESS |
204 | #define ORANGEFS_XATTR_NAME_ACL_DEFAULT POSIX_ACL_XATTR_DEFAULT | |
205 | #define ORANGEFS_XATTR_NAME_TRUSTED_PREFIX "trusted." | |
206 | #define ORANGEFS_XATTR_NAME_DEFAULT_PREFIX "" | |
f7ab093f | 207 | |
8bb8aefd | 208 | /* these functions are defined in orangefs-utils.c */ |
f7ab093f MM |
209 | int orangefs_prepare_cdm_array(char *debug_array_string); |
210 | int orangefs_prepare_debugfs_help_string(int); | |
211 | ||
8bb8aefd YL |
212 | /* defined in orangefs-debugfs.c */ |
213 | int orangefs_client_debug_init(void); | |
f7ab093f MM |
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 | ||
8bb8aefd YL |
225 | /*these variables are defined in orangefs-mod.c */ |
226 | extern char kernel_debug_string[ORANGEFS_MAX_DEBUG_STRING_LEN]; | |
227 | extern char client_debug_string[ORANGEFS_MAX_DEBUG_STRING_LEN]; | |
228 | extern char client_debug_array_string[ORANGEFS_MAX_DEBUG_STRING_LEN]; | |
f7ab093f MM |
229 | extern unsigned int kernel_mask_set_mod_init; |
230 | ||
8bb8aefd YL |
231 | extern int orangefs_init_acl(struct inode *inode, struct inode *dir); |
232 | extern const struct xattr_handler *orangefs_xattr_handlers[]; | |
f7ab093f | 233 | |
8bb8aefd YL |
234 | extern struct posix_acl *orangefs_get_acl(struct inode *inode, int type); |
235 | extern int orangefs_set_acl(struct inode *inode, struct posix_acl *acl, int type); | |
f7ab093f | 236 | |
f7ab093f MM |
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 | /* | |
8bb8aefd | 247 | * orangefs data structures |
f7ab093f | 248 | */ |
8bb8aefd YL |
249 | struct orangefs_kernel_op_s { |
250 | enum orangefs_vfs_op_states op_state; | |
f7ab093f MM |
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 | ||
8bb8aefd YL |
260 | struct orangefs_upcall_s upcall; |
261 | struct orangefs_downcall_s downcall; | |
f7ab093f MM |
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 | ||
f7ab093f MM |
269 | /* VFS aio fields */ |
270 | ||
8bb8aefd | 271 | /* used by the async I/O code to stash the orangefs_kiocb_s structure */ |
f7ab093f MM |
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 | ||
8bb8aefd YL |
282 | /* per inode private orangefs info */ |
283 | struct orangefs_inode_s { | |
284 | struct orangefs_object_kref refn; | |
285 | char link_target[ORANGEFS_NAME_MAX]; | |
f7ab093f MM |
286 | __s64 blksize; |
287 | /* | |
288 | * Reading/Writing Extended attributes need to acquire the appropriate | |
8bb8aefd | 289 | * reader/writer semaphore on the orangefs_inode_s structure. |
f7ab093f MM |
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 | ||
8bb8aefd | 302 | /* All allocated orangefs_inode_s objects are chained to a list */ |
f7ab093f MM |
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 | ||
8bb8aefd YL |
327 | /* per superblock private orangefs info */ |
328 | struct orangefs_sb_info_s { | |
329 | struct orangefs_khandle root_khandle; | |
f7ab093f MM |
330 | __s32 fs_id; |
331 | int id; | |
332 | int flags; | |
8bb8aefd YL |
333 | #define ORANGEFS_OPT_INTR 0x01 |
334 | #define ORANGEFS_OPT_LOCAL_LOCK 0x02 | |
335 | char devname[ORANGEFS_MAX_SERVER_ADDR_LEN]; | |
f7ab093f MM |
336 | struct super_block *sb; |
337 | int mount_pending; | |
338 | struct list_head list; | |
339 | }; | |
340 | ||
f7ab093f MM |
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 | */ | |
8bb8aefd | 347 | struct orangefs_kiocb_s { |
f7ab093f MM |
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 */ | |
8bb8aefd | 355 | struct orangefs_bufmap *bufmap; |
f7ab093f MM |
356 | int buffer_index; |
357 | ||
8bb8aefd YL |
358 | /* orangefs kernel operation type */ |
359 | struct orangefs_kernel_op_s *op; | |
f7ab093f MM |
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 | ||
8bb8aefd | 380 | struct orangefs_stats { |
f7ab093f MM |
381 | unsigned long cache_hits; |
382 | unsigned long cache_misses; | |
383 | unsigned long reads; | |
384 | unsigned long writes; | |
385 | }; | |
386 | ||
8bb8aefd | 387 | extern struct orangefs_stats g_orangefs_stats; |
f7ab093f MM |
388 | |
389 | /* | |
54804949 MM |
390 | * NOTE: See Documentation/filesystems/porting for information |
391 | * on implementing FOO_I and properly accessing fs private data | |
392 | */ | |
8bb8aefd | 393 | static inline struct orangefs_inode_s *ORANGEFS_I(struct inode *inode) |
f7ab093f | 394 | { |
8bb8aefd | 395 | return container_of(inode, struct orangefs_inode_s, vfs_inode); |
f7ab093f MM |
396 | } |
397 | ||
8bb8aefd | 398 | static inline struct orangefs_sb_info_s *ORANGEFS_SB(struct super_block *sb) |
f7ab093f | 399 | { |
8bb8aefd | 400 | return (struct orangefs_sb_info_s *) sb->s_fs_info; |
f7ab093f MM |
401 | } |
402 | ||
403 | /* ino_t descends from "unsigned long", 8 bytes, 64 bits. */ | |
8bb8aefd | 404 | static inline ino_t orangefs_khandle_to_ino(struct orangefs_khandle *khandle) |
f7ab093f MM |
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 | ||
8bb8aefd | 423 | static inline struct orangefs_khandle *get_khandle_from_ino(struct inode *inode) |
f7ab093f | 424 | { |
8bb8aefd | 425 | return &(ORANGEFS_I(inode)->refn.khandle); |
f7ab093f MM |
426 | } |
427 | ||
428 | static inline __s32 get_fsid_from_ino(struct inode *inode) | |
429 | { | |
8bb8aefd | 430 | return ORANGEFS_I(inode)->refn.fs_id; |
f7ab093f MM |
431 | } |
432 | ||
433 | static inline ino_t get_ino_from_khandle(struct inode *inode) | |
434 | { | |
8bb8aefd | 435 | struct orangefs_khandle *khandle; |
f7ab093f MM |
436 | ino_t ino; |
437 | ||
438 | khandle = get_khandle_from_ino(inode); | |
8bb8aefd | 439 | ino = orangefs_khandle_to_ino(khandle); |
f7ab093f MM |
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__, | |
8bb8aefd | 453 | &ORANGEFS_SB(inode->i_sb)->root_khandle, |
f7ab093f MM |
454 | get_khandle_from_ino(inode)); |
455 | ||
8bb8aefd | 456 | if (ORANGEFS_khandle_cmp(&(ORANGEFS_SB(inode->i_sb)->root_khandle), |
f7ab093f MM |
457 | get_khandle_from_ino(inode))) |
458 | return 0; | |
459 | else | |
460 | return 1; | |
461 | } | |
462 | ||
8bb8aefd | 463 | static inline int match_handle(struct orangefs_khandle resp_handle, |
f7ab093f MM |
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 | ||
8bb8aefd | 472 | if (ORANGEFS_khandle_cmp(&resp_handle, get_khandle_from_ino(inode))) |
f7ab093f MM |
473 | return 0; |
474 | else | |
475 | return 1; | |
476 | } | |
477 | ||
478 | /* | |
8bb8aefd | 479 | * defined in orangefs-cache.c |
f7ab093f MM |
480 | */ |
481 | int op_cache_initialize(void); | |
482 | int op_cache_finalize(void); | |
8bb8aefd YL |
483 | struct orangefs_kernel_op_s *op_alloc(__s32 type); |
484 | char *get_opname_string(struct orangefs_kernel_op_s *new_op); | |
485 | void op_release(struct orangefs_kernel_op_s *op); | |
f7ab093f MM |
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 | ||
8bb8aefd YL |
492 | int orangefs_inode_cache_initialize(void); |
493 | int orangefs_inode_cache_finalize(void); | |
f7ab093f MM |
494 | |
495 | int kiocb_cache_initialize(void); | |
496 | int kiocb_cache_finalize(void); | |
8bb8aefd YL |
497 | struct orangefs_kiocb_s *kiocb_alloc(void); |
498 | void kiocb_release(struct orangefs_kiocb_s *ptr); | |
f7ab093f MM |
499 | |
500 | /* | |
8bb8aefd | 501 | * defined in orangefs-mod.c |
f7ab093f MM |
502 | */ |
503 | void purge_inprogress_ops(void); | |
504 | ||
505 | /* | |
506 | * defined in waitqueue.c | |
507 | */ | |
8bb8aefd YL |
508 | int wait_for_matching_downcall(struct orangefs_kernel_op_s *op); |
509 | int wait_for_cancellation_downcall(struct orangefs_kernel_op_s *op); | |
510 | void orangefs_clean_up_interrupted_operation(struct orangefs_kernel_op_s *op); | |
f7ab093f MM |
511 | void purge_waiting_ops(void); |
512 | ||
513 | /* | |
514 | * defined in super.c | |
515 | */ | |
8bb8aefd | 516 | struct dentry *orangefs_mount(struct file_system_type *fst, |
f7ab093f MM |
517 | int flags, |
518 | const char *devname, | |
519 | void *data); | |
520 | ||
8bb8aefd YL |
521 | void orangefs_kill_sb(struct super_block *sb); |
522 | int orangefs_remount(struct super_block *sb); | |
f7ab093f MM |
523 | |
524 | int fsid_key_table_initialize(void); | |
525 | void fsid_key_table_finalize(void); | |
526 | ||
527 | /* | |
528 | * defined in inode.c | |
529 | */ | |
8bb8aefd YL |
530 | __u32 convert_to_orangefs_mask(unsigned long lite_mask); |
531 | struct inode *orangefs_new_inode(struct super_block *sb, | |
f7ab093f MM |
532 | struct inode *dir, |
533 | int mode, | |
534 | dev_t dev, | |
8bb8aefd | 535 | struct orangefs_object_kref *ref); |
f7ab093f | 536 | |
8bb8aefd | 537 | int orangefs_setattr(struct dentry *dentry, struct iattr *iattr); |
f7ab093f | 538 | |
8bb8aefd | 539 | int orangefs_getattr(struct vfsmount *mnt, |
f7ab093f MM |
540 | struct dentry *dentry, |
541 | struct kstat *kstat); | |
542 | ||
543 | /* | |
544 | * defined in xattr.c | |
545 | */ | |
8bb8aefd | 546 | int orangefs_setxattr(struct dentry *dentry, |
f7ab093f MM |
547 | const char *name, |
548 | const void *value, | |
549 | size_t size, | |
550 | int flags); | |
551 | ||
8bb8aefd | 552 | ssize_t orangefs_getxattr(struct dentry *dentry, |
f7ab093f MM |
553 | const char *name, |
554 | void *buffer, | |
555 | size_t size); | |
556 | ||
8bb8aefd | 557 | ssize_t orangefs_listxattr(struct dentry *dentry, char *buffer, size_t size); |
f7ab093f MM |
558 | |
559 | /* | |
560 | * defined in namei.c | |
561 | */ | |
8bb8aefd YL |
562 | struct inode *orangefs_iget(struct super_block *sb, |
563 | struct orangefs_object_kref *ref); | |
f7ab093f | 564 | |
8bb8aefd YL |
565 | ssize_t orangefs_inode_read(struct inode *inode, |
566 | struct iov_iter *iter, | |
567 | loff_t *offset, | |
568 | loff_t readahead_size); | |
f7ab093f MM |
569 | |
570 | /* | |
8bb8aefd | 571 | * defined in devorangefs-req.c |
f7ab093f | 572 | */ |
8bb8aefd YL |
573 | int orangefs_dev_init(void); |
574 | void orangefs_dev_cleanup(void); | |
f7ab093f MM |
575 | int is_daemon_in_service(void); |
576 | int fs_mount_pending(__s32 fsid); | |
577 | ||
578 | /* | |
8bb8aefd | 579 | * defined in orangefs-utils.c |
f7ab093f | 580 | */ |
8bb8aefd | 581 | __s32 fsid_of_op(struct orangefs_kernel_op_s *op); |
f7ab093f | 582 | |
8bb8aefd | 583 | int orangefs_flush_inode(struct inode *inode); |
f7ab093f | 584 | |
8bb8aefd | 585 | ssize_t orangefs_inode_getxattr(struct inode *inode, |
f7ab093f MM |
586 | const char *prefix, |
587 | const char *name, | |
588 | void *buffer, | |
589 | size_t size); | |
590 | ||
8bb8aefd | 591 | int orangefs_inode_setxattr(struct inode *inode, |
f7ab093f MM |
592 | const char *prefix, |
593 | const char *name, | |
594 | const void *value, | |
595 | size_t size, | |
596 | int flags); | |
597 | ||
8bb8aefd | 598 | int orangefs_inode_getattr(struct inode *inode, __u32 mask); |
f7ab093f | 599 | |
8bb8aefd | 600 | int orangefs_inode_setattr(struct inode *inode, struct iattr *iattr); |
f7ab093f | 601 | |
8bb8aefd | 602 | void orangefs_op_initialize(struct orangefs_kernel_op_s *op); |
f7ab093f | 603 | |
8bb8aefd | 604 | void orangefs_make_bad_inode(struct inode *inode); |
f7ab093f | 605 | |
8c3905ad | 606 | void block_signals(sigset_t *); |
f7ab093f | 607 | |
8c3905ad | 608 | void set_signals(sigset_t *); |
f7ab093f | 609 | |
8bb8aefd | 610 | int orangefs_unmount_sb(struct super_block *sb); |
f7ab093f | 611 | |
8bb8aefd | 612 | int orangefs_cancel_op_in_progress(__u64 tag); |
f7ab093f | 613 | |
8bb8aefd | 614 | static inline __u64 orangefs_convert_time_field(const struct timespec *ts) |
5714156b AV |
615 | { |
616 | return (__u64)ts->tv_sec; | |
617 | } | |
f7ab093f | 618 | |
8bb8aefd | 619 | int orangefs_normalize_to_errno(__s32 error_code); |
f7ab093f MM |
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; | |
8bb8aefd YL |
626 | extern struct list_head orangefs_superblocks; |
627 | extern spinlock_t orangefs_superblocks_lock; | |
628 | extern struct list_head orangefs_request_list; | |
629 | extern spinlock_t orangefs_request_list_lock; | |
630 | extern wait_queue_head_t orangefs_request_list_waitq; | |
f7ab093f MM |
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 | ||
8bb8aefd YL |
635 | extern const struct address_space_operations orangefs_address_operations; |
636 | extern struct backing_dev_info orangefs_backing_dev_info; | |
637 | extern struct inode_operations orangefs_file_inode_operations; | |
638 | extern const struct file_operations orangefs_file_operations; | |
639 | extern struct inode_operations orangefs_symlink_inode_operations; | |
640 | extern struct inode_operations orangefs_dir_inode_operations; | |
641 | extern const struct file_operations orangefs_dir_operations; | |
642 | extern const struct dentry_operations orangefs_dentry_operations; | |
643 | extern const struct file_operations orangefs_devreq_file_operations; | |
f7ab093f | 644 | |
8bb8aefd | 645 | extern wait_queue_head_t orangefs_bufmap_init_waitq; |
f7ab093f MM |
646 | |
647 | /* | |
648 | * misc convenience macros | |
649 | */ | |
650 | #define add_op_to_request_list(op) \ | |
651 | do { \ | |
8bb8aefd | 652 | spin_lock(&orangefs_request_list_lock); \ |
f7ab093f MM |
653 | spin_lock(&op->lock); \ |
654 | set_op_state_waiting(op); \ | |
8bb8aefd YL |
655 | list_add_tail(&op->list, &orangefs_request_list); \ |
656 | spin_unlock(&orangefs_request_list_lock); \ | |
f7ab093f | 657 | spin_unlock(&op->lock); \ |
8bb8aefd | 658 | wake_up_interruptible(&orangefs_request_list_waitq); \ |
f7ab093f MM |
659 | } while (0) |
660 | ||
661 | #define add_priority_op_to_request_list(op) \ | |
662 | do { \ | |
8bb8aefd | 663 | spin_lock(&orangefs_request_list_lock); \ |
f7ab093f MM |
664 | spin_lock(&op->lock); \ |
665 | set_op_state_waiting(op); \ | |
666 | \ | |
8bb8aefd YL |
667 | list_add(&op->list, &orangefs_request_list); \ |
668 | spin_unlock(&orangefs_request_list_lock); \ | |
f7ab093f | 669 | spin_unlock(&op->lock); \ |
8bb8aefd | 670 | wake_up_interruptible(&orangefs_request_list_waitq); \ |
f7ab093f MM |
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; \ | |
8bb8aefd | 677 | struct orangefs_kernel_op_s *tmp_op = NULL; \ |
f7ab093f | 678 | \ |
8bb8aefd YL |
679 | spin_lock(&orangefs_request_list_lock); \ |
680 | list_for_each_safe(tmp, tmp_safe, &orangefs_request_list) { \ | |
f7ab093f | 681 | tmp_op = list_entry(tmp, \ |
8bb8aefd | 682 | struct orangefs_kernel_op_s, \ |
f7ab093f MM |
683 | list); \ |
684 | if (tmp_op && (tmp_op == op)) { \ | |
685 | list_del(&tmp_op->list); \ | |
686 | break; \ | |
687 | } \ | |
688 | } \ | |
8bb8aefd | 689 | spin_unlock(&orangefs_request_list_lock); \ |
f7ab093f MM |
690 | } while (0) |
691 | ||
8bb8aefd YL |
692 | #define ORANGEFS_OP_INTERRUPTIBLE 1 /* service_operation() is interruptible */ |
693 | #define ORANGEFS_OP_PRIORITY 2 /* service_operation() is high priority */ | |
694 | #define ORANGEFS_OP_CANCELLATION 4 /* this is a cancellation */ | |
695 | #define ORANGEFS_OP_NO_SEMAPHORE 8 /* don't acquire semaphore */ | |
696 | #define ORANGEFS_OP_ASYNC 16 /* Queue it, but don't wait */ | |
f7ab093f | 697 | |
8bb8aefd | 698 | int service_operation(struct orangefs_kernel_op_s *op, |
f7ab093f MM |
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 | * | |
8bb8aefd | 722 | * if a orangefs sysint level error occured and i/o has been completed, |
f7ab093f MM |
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)) { \ | |
8bb8aefd | 734 | orangefs_cancel_op_in_progress(new_op->tag); \ |
f7ab093f MM |
735 | op_release(new_op); \ |
736 | } else { \ | |
737 | wake_up_daemon_for_return(new_op); \ | |
738 | } \ | |
739 | new_op = NULL; \ | |
8bb8aefd | 740 | orangefs_bufmap_put(bufmap, buffer_index); \ |
f7ab093f MM |
741 | buffer_index = -1; \ |
742 | } while (0) | |
743 | ||
744 | #define get_interruptible_flag(inode) \ | |
8bb8aefd YL |
745 | ((ORANGEFS_SB(inode->i_sb)->flags & ORANGEFS_OPT_INTR) ? \ |
746 | ORANGEFS_OP_INTERRUPTIBLE : 0) | |
f7ab093f | 747 | |
8bb8aefd | 748 | #define add_orangefs_sb(sb) \ |
f7ab093f MM |
749 | do { \ |
750 | gossip_debug(GOSSIP_SUPER_DEBUG, \ | |
8bb8aefd YL |
751 | "Adding SB %p to orangefs superblocks\n", \ |
752 | ORANGEFS_SB(sb)); \ | |
753 | spin_lock(&orangefs_superblocks_lock); \ | |
754 | list_add_tail(&ORANGEFS_SB(sb)->list, &orangefs_superblocks); \ | |
755 | spin_unlock(&orangefs_superblocks_lock); \ | |
f7ab093f MM |
756 | } while (0) |
757 | ||
8bb8aefd | 758 | #define remove_orangefs_sb(sb) \ |
f7ab093f MM |
759 | do { \ |
760 | struct list_head *tmp = NULL; \ | |
761 | struct list_head *tmp_safe = NULL; \ | |
8bb8aefd | 762 | struct orangefs_sb_info_s *orangefs_sb = NULL; \ |
f7ab093f | 763 | \ |
8bb8aefd YL |
764 | spin_lock(&orangefs_superblocks_lock); \ |
765 | list_for_each_safe(tmp, tmp_safe, &orangefs_superblocks) { \ | |
766 | orangefs_sb = list_entry(tmp, \ | |
767 | struct orangefs_sb_info_s, \ | |
f7ab093f | 768 | list); \ |
8bb8aefd | 769 | if (orangefs_sb && (orangefs_sb->sb == sb)) { \ |
f7ab093f | 770 | gossip_debug(GOSSIP_SUPER_DEBUG, \ |
8bb8aefd YL |
771 | "Removing SB %p from orangefs superblocks\n", \ |
772 | orangefs_sb); \ | |
773 | list_del(&orangefs_sb->list); \ | |
f7ab093f MM |
774 | break; \ |
775 | } \ | |
776 | } \ | |
8bb8aefd | 777 | spin_unlock(&orangefs_superblocks_lock); \ |
f7ab093f MM |
778 | } while (0) |
779 | ||
8bb8aefd YL |
780 | #define orangefs_lock_inode(inode) spin_lock(&inode->i_lock) |
781 | #define orangefs_unlock_inode(inode) spin_unlock(&inode->i_lock) | |
f7ab093f MM |
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; \ | |
8bb8aefd | 788 | sys_attr.perms = ORANGEFS_util_translate_mode(mode); \ |
f7ab093f | 789 | sys_attr.objtype = type; \ |
8bb8aefd | 790 | sys_attr.mask = ORANGEFS_ATTR_SYS_ALL_SETABLE; \ |
f7ab093f MM |
791 | } while (0) |
792 | ||
8bb8aefd | 793 | #define orangefs_inode_lock(__i) mutex_lock(&(__i)->i_mutex) |
f7ab093f | 794 | |
8bb8aefd | 795 | #define orangefs_inode_unlock(__i) mutex_unlock(&(__i)->i_mutex) |
f7ab093f | 796 | |
8bb8aefd | 797 | static inline void orangefs_i_size_write(struct inode *inode, loff_t i_size) |
f7ab093f MM |
798 | { |
799 | #if BITS_PER_LONG == 32 && defined(CONFIG_SMP) | |
8bb8aefd | 800 | ornagefs_inode_lock(inode); |
f7ab093f MM |
801 | #endif |
802 | i_size_write(inode, i_size); | |
803 | #if BITS_PER_LONG == 32 && defined(CONFIG_SMP) | |
8bb8aefd | 804 | orangefs_inode_unlock(inode); |
f7ab093f MM |
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 | ||
8bb8aefd | 819 | #endif /* __ORANGEFSKERNEL_H */ |