Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1 | /* |
| 2 | * linux/fs/locks.c |
| 3 | * |
| 4 | * Provide support for fcntl()'s F_GETLK, F_SETLK, and F_SETLKW calls. |
| 5 | * Doug Evans (dje@spiff.uucp), August 07, 1992 |
| 6 | * |
| 7 | * Deadlock detection added. |
| 8 | * FIXME: one thing isn't handled yet: |
| 9 | * - mandatory locks (requires lots of changes elsewhere) |
| 10 | * Kelly Carmichael (kelly@[142.24.8.65]), September 17, 1994. |
| 11 | * |
| 12 | * Miscellaneous edits, and a total rewrite of posix_lock_file() code. |
| 13 | * Kai Petzke (wpp@marie.physik.tu-berlin.de), 1994 |
| 14 | * |
| 15 | * Converted file_lock_table to a linked list from an array, which eliminates |
| 16 | * the limits on how many active file locks are open. |
| 17 | * Chad Page (pageone@netcom.com), November 27, 1994 |
| 18 | * |
| 19 | * Removed dependency on file descriptors. dup()'ed file descriptors now |
| 20 | * get the same locks as the original file descriptors, and a close() on |
| 21 | * any file descriptor removes ALL the locks on the file for the current |
| 22 | * process. Since locks still depend on the process id, locks are inherited |
| 23 | * after an exec() but not after a fork(). This agrees with POSIX, and both |
| 24 | * BSD and SVR4 practice. |
| 25 | * Andy Walker (andy@lysaker.kvaerner.no), February 14, 1995 |
| 26 | * |
| 27 | * Scrapped free list which is redundant now that we allocate locks |
| 28 | * dynamically with kmalloc()/kfree(). |
| 29 | * Andy Walker (andy@lysaker.kvaerner.no), February 21, 1995 |
| 30 | * |
| 31 | * Implemented two lock personalities - FL_FLOCK and FL_POSIX. |
| 32 | * |
| 33 | * FL_POSIX locks are created with calls to fcntl() and lockf() through the |
| 34 | * fcntl() system call. They have the semantics described above. |
| 35 | * |
| 36 | * FL_FLOCK locks are created with calls to flock(), through the flock() |
| 37 | * system call, which is new. Old C libraries implement flock() via fcntl() |
| 38 | * and will continue to use the old, broken implementation. |
| 39 | * |
| 40 | * FL_FLOCK locks follow the 4.4 BSD flock() semantics. They are associated |
| 41 | * with a file pointer (filp). As a result they can be shared by a parent |
| 42 | * process and its children after a fork(). They are removed when the last |
| 43 | * file descriptor referring to the file pointer is closed (unless explicitly |
| 44 | * unlocked). |
| 45 | * |
| 46 | * FL_FLOCK locks never deadlock, an existing lock is always removed before |
| 47 | * upgrading from shared to exclusive (or vice versa). When this happens |
| 48 | * any processes blocked by the current lock are woken up and allowed to |
| 49 | * run before the new lock is applied. |
| 50 | * Andy Walker (andy@lysaker.kvaerner.no), June 09, 1995 |
| 51 | * |
| 52 | * Removed some race conditions in flock_lock_file(), marked other possible |
| 53 | * races. Just grep for FIXME to see them. |
| 54 | * Dmitry Gorodchanin (pgmdsg@ibi.com), February 09, 1996. |
| 55 | * |
| 56 | * Addressed Dmitry's concerns. Deadlock checking no longer recursive. |
| 57 | * Lock allocation changed to GFP_ATOMIC as we can't afford to sleep |
| 58 | * once we've checked for blocking and deadlocking. |
| 59 | * Andy Walker (andy@lysaker.kvaerner.no), April 03, 1996. |
| 60 | * |
| 61 | * Initial implementation of mandatory locks. SunOS turned out to be |
| 62 | * a rotten model, so I implemented the "obvious" semantics. |
| 63 | * See 'Documentation/mandatory.txt' for details. |
| 64 | * Andy Walker (andy@lysaker.kvaerner.no), April 06, 1996. |
| 65 | * |
| 66 | * Don't allow mandatory locks on mmap()'ed files. Added simple functions to |
| 67 | * check if a file has mandatory locks, used by mmap(), open() and creat() to |
| 68 | * see if system call should be rejected. Ref. HP-UX/SunOS/Solaris Reference |
| 69 | * Manual, Section 2. |
| 70 | * Andy Walker (andy@lysaker.kvaerner.no), April 09, 1996. |
| 71 | * |
| 72 | * Tidied up block list handling. Added '/proc/locks' interface. |
| 73 | * Andy Walker (andy@lysaker.kvaerner.no), April 24, 1996. |
| 74 | * |
| 75 | * Fixed deadlock condition for pathological code that mixes calls to |
| 76 | * flock() and fcntl(). |
| 77 | * Andy Walker (andy@lysaker.kvaerner.no), April 29, 1996. |
| 78 | * |
| 79 | * Allow only one type of locking scheme (FL_POSIX or FL_FLOCK) to be in use |
| 80 | * for a given file at a time. Changed the CONFIG_LOCK_MANDATORY scheme to |
| 81 | * guarantee sensible behaviour in the case where file system modules might |
| 82 | * be compiled with different options than the kernel itself. |
| 83 | * Andy Walker (andy@lysaker.kvaerner.no), May 15, 1996. |
| 84 | * |
| 85 | * Added a couple of missing wake_up() calls. Thanks to Thomas Meckel |
| 86 | * (Thomas.Meckel@mni.fh-giessen.de) for spotting this. |
| 87 | * Andy Walker (andy@lysaker.kvaerner.no), May 15, 1996. |
| 88 | * |
| 89 | * Changed FL_POSIX locks to use the block list in the same way as FL_FLOCK |
| 90 | * locks. Changed process synchronisation to avoid dereferencing locks that |
| 91 | * have already been freed. |
| 92 | * Andy Walker (andy@lysaker.kvaerner.no), Sep 21, 1996. |
| 93 | * |
| 94 | * Made the block list a circular list to minimise searching in the list. |
| 95 | * Andy Walker (andy@lysaker.kvaerner.no), Sep 25, 1996. |
| 96 | * |
| 97 | * Made mandatory locking a mount option. Default is not to allow mandatory |
| 98 | * locking. |
| 99 | * Andy Walker (andy@lysaker.kvaerner.no), Oct 04, 1996. |
| 100 | * |
| 101 | * Some adaptations for NFS support. |
| 102 | * Olaf Kirch (okir@monad.swb.de), Dec 1996, |
| 103 | * |
| 104 | * Fixed /proc/locks interface so that we can't overrun the buffer we are handed. |
| 105 | * Andy Walker (andy@lysaker.kvaerner.no), May 12, 1997. |
| 106 | * |
| 107 | * Use slab allocator instead of kmalloc/kfree. |
| 108 | * Use generic list implementation from <linux/list.h>. |
| 109 | * Sped up posix_locks_deadlock by only considering blocked locks. |
| 110 | * Matthew Wilcox <willy@debian.org>, March, 2000. |
| 111 | * |
| 112 | * Leases and LOCK_MAND |
| 113 | * Matthew Wilcox <willy@debian.org>, June, 2000. |
| 114 | * Stephen Rothwell <sfr@canb.auug.org.au>, June, 2000. |
| 115 | */ |
| 116 | |
| 117 | #include <linux/capability.h> |
| 118 | #include <linux/file.h> |
| 119 | #include <linux/fs.h> |
| 120 | #include <linux/init.h> |
| 121 | #include <linux/module.h> |
| 122 | #include <linux/security.h> |
| 123 | #include <linux/slab.h> |
| 124 | #include <linux/smp_lock.h> |
| 125 | #include <linux/syscalls.h> |
| 126 | #include <linux/time.h> |
Dipankar Sarma | 4fb3a53 | 2005-09-16 19:28:13 -0700 | [diff] [blame] | 127 | #include <linux/rcupdate.h> |
Vitaliy Gusev | ab1f161 | 2008-01-17 00:07:08 +0000 | [diff] [blame] | 128 | #include <linux/pid_namespace.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 129 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 130 | #include <asm/uaccess.h> |
| 131 | |
| 132 | #define IS_POSIX(fl) (fl->fl_flags & FL_POSIX) |
| 133 | #define IS_FLOCK(fl) (fl->fl_flags & FL_FLOCK) |
| 134 | #define IS_LEASE(fl) (fl->fl_flags & FL_LEASE) |
| 135 | |
| 136 | int leases_enable = 1; |
| 137 | int lease_break_time = 45; |
| 138 | |
| 139 | #define for_each_lock(inode, lockp) \ |
| 140 | for (lockp = &inode->i_flock; *lockp != NULL; lockp = &(*lockp)->fl_next) |
| 141 | |
Christoph Hellwig | 26bcbf9 | 2006-03-20 13:44:40 -0500 | [diff] [blame] | 142 | static LIST_HEAD(file_lock_list); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 143 | static LIST_HEAD(blocked_list); |
| 144 | |
Christoph Lameter | e18b890 | 2006-12-06 20:33:20 -0800 | [diff] [blame] | 145 | static struct kmem_cache *filelock_cache __read_mostly; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 146 | |
| 147 | /* Allocate an empty lock structure. */ |
| 148 | static struct file_lock *locks_alloc_lock(void) |
| 149 | { |
Christoph Lameter | e94b176 | 2006-12-06 20:33:17 -0800 | [diff] [blame] | 150 | return kmem_cache_alloc(filelock_cache, GFP_KERNEL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 151 | } |
| 152 | |
Trond Myklebust | 47831f3 | 2006-03-20 13:44:05 -0500 | [diff] [blame] | 153 | static void locks_release_private(struct file_lock *fl) |
| 154 | { |
| 155 | if (fl->fl_ops) { |
| 156 | if (fl->fl_ops->fl_release_private) |
| 157 | fl->fl_ops->fl_release_private(fl); |
| 158 | fl->fl_ops = NULL; |
| 159 | } |
| 160 | if (fl->fl_lmops) { |
| 161 | if (fl->fl_lmops->fl_release_private) |
| 162 | fl->fl_lmops->fl_release_private(fl); |
| 163 | fl->fl_lmops = NULL; |
| 164 | } |
| 165 | |
| 166 | } |
| 167 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 168 | /* Free a lock which is not in use. */ |
Matt Mackall | 33443c4 | 2006-01-08 01:05:22 -0800 | [diff] [blame] | 169 | static void locks_free_lock(struct file_lock *fl) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 170 | { |
Miklos Szeredi | 5ce2964 | 2006-03-31 02:30:29 -0800 | [diff] [blame] | 171 | BUG_ON(waitqueue_active(&fl->fl_wait)); |
| 172 | BUG_ON(!list_empty(&fl->fl_block)); |
| 173 | BUG_ON(!list_empty(&fl->fl_link)); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 174 | |
Trond Myklebust | 47831f3 | 2006-03-20 13:44:05 -0500 | [diff] [blame] | 175 | locks_release_private(fl); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 176 | kmem_cache_free(filelock_cache, fl); |
| 177 | } |
| 178 | |
| 179 | void locks_init_lock(struct file_lock *fl) |
| 180 | { |
| 181 | INIT_LIST_HEAD(&fl->fl_link); |
| 182 | INIT_LIST_HEAD(&fl->fl_block); |
| 183 | init_waitqueue_head(&fl->fl_wait); |
| 184 | fl->fl_next = NULL; |
| 185 | fl->fl_fasync = NULL; |
| 186 | fl->fl_owner = NULL; |
| 187 | fl->fl_pid = 0; |
Vitaliy Gusev | ab1f161 | 2008-01-17 00:07:08 +0000 | [diff] [blame] | 188 | fl->fl_nspid = NULL; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 189 | fl->fl_file = NULL; |
| 190 | fl->fl_flags = 0; |
| 191 | fl->fl_type = 0; |
| 192 | fl->fl_start = fl->fl_end = 0; |
| 193 | fl->fl_ops = NULL; |
| 194 | fl->fl_lmops = NULL; |
| 195 | } |
| 196 | |
| 197 | EXPORT_SYMBOL(locks_init_lock); |
| 198 | |
| 199 | /* |
| 200 | * Initialises the fields of the file lock which are invariant for |
| 201 | * free file_locks. |
| 202 | */ |
Christoph Lameter | 4ba9b9d | 2007-10-16 23:25:51 -0700 | [diff] [blame] | 203 | static void init_once(struct kmem_cache *cache, void *foo) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 204 | { |
| 205 | struct file_lock *lock = (struct file_lock *) foo; |
| 206 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 207 | locks_init_lock(lock); |
| 208 | } |
| 209 | |
Trond Myklebust | 47831f3 | 2006-03-20 13:44:05 -0500 | [diff] [blame] | 210 | static void locks_copy_private(struct file_lock *new, struct file_lock *fl) |
| 211 | { |
| 212 | if (fl->fl_ops) { |
| 213 | if (fl->fl_ops->fl_copy_lock) |
| 214 | fl->fl_ops->fl_copy_lock(new, fl); |
| 215 | new->fl_ops = fl->fl_ops; |
| 216 | } |
| 217 | if (fl->fl_lmops) { |
| 218 | if (fl->fl_lmops->fl_copy_lock) |
| 219 | fl->fl_lmops->fl_copy_lock(new, fl); |
| 220 | new->fl_lmops = fl->fl_lmops; |
| 221 | } |
| 222 | } |
| 223 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 224 | /* |
| 225 | * Initialize a new lock from an existing file_lock structure. |
| 226 | */ |
J. Bruce Fields | 1a747ee | 2008-04-24 10:08:22 -0400 | [diff] [blame] | 227 | void __locks_copy_lock(struct file_lock *new, const struct file_lock *fl) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 228 | { |
| 229 | new->fl_owner = fl->fl_owner; |
| 230 | new->fl_pid = fl->fl_pid; |
Trond Myklebust | 0996905 | 2006-03-20 13:44:38 -0500 | [diff] [blame] | 231 | new->fl_file = NULL; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 232 | new->fl_flags = fl->fl_flags; |
| 233 | new->fl_type = fl->fl_type; |
| 234 | new->fl_start = fl->fl_start; |
| 235 | new->fl_end = fl->fl_end; |
Trond Myklebust | 0996905 | 2006-03-20 13:44:38 -0500 | [diff] [blame] | 236 | new->fl_ops = NULL; |
| 237 | new->fl_lmops = NULL; |
| 238 | } |
| 239 | |
| 240 | void locks_copy_lock(struct file_lock *new, struct file_lock *fl) |
| 241 | { |
| 242 | locks_release_private(new); |
| 243 | |
| 244 | __locks_copy_lock(new, fl); |
| 245 | new->fl_file = fl->fl_file; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 246 | new->fl_ops = fl->fl_ops; |
| 247 | new->fl_lmops = fl->fl_lmops; |
Trond Myklebust | 47831f3 | 2006-03-20 13:44:05 -0500 | [diff] [blame] | 248 | |
| 249 | locks_copy_private(new, fl); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 250 | } |
| 251 | |
| 252 | EXPORT_SYMBOL(locks_copy_lock); |
| 253 | |
| 254 | static inline int flock_translate_cmd(int cmd) { |
| 255 | if (cmd & LOCK_MAND) |
| 256 | return cmd & (LOCK_MAND | LOCK_RW); |
| 257 | switch (cmd) { |
| 258 | case LOCK_SH: |
| 259 | return F_RDLCK; |
| 260 | case LOCK_EX: |
| 261 | return F_WRLCK; |
| 262 | case LOCK_UN: |
| 263 | return F_UNLCK; |
| 264 | } |
| 265 | return -EINVAL; |
| 266 | } |
| 267 | |
| 268 | /* Fill in a file_lock structure with an appropriate FLOCK lock. */ |
| 269 | static int flock_make_lock(struct file *filp, struct file_lock **lock, |
| 270 | unsigned int cmd) |
| 271 | { |
| 272 | struct file_lock *fl; |
| 273 | int type = flock_translate_cmd(cmd); |
| 274 | if (type < 0) |
| 275 | return type; |
| 276 | |
| 277 | fl = locks_alloc_lock(); |
| 278 | if (fl == NULL) |
| 279 | return -ENOMEM; |
| 280 | |
| 281 | fl->fl_file = filp; |
| 282 | fl->fl_pid = current->tgid; |
| 283 | fl->fl_flags = FL_FLOCK; |
| 284 | fl->fl_type = type; |
| 285 | fl->fl_end = OFFSET_MAX; |
| 286 | |
| 287 | *lock = fl; |
| 288 | return 0; |
| 289 | } |
| 290 | |
| 291 | static int assign_type(struct file_lock *fl, int type) |
| 292 | { |
| 293 | switch (type) { |
| 294 | case F_RDLCK: |
| 295 | case F_WRLCK: |
| 296 | case F_UNLCK: |
| 297 | fl->fl_type = type; |
| 298 | break; |
| 299 | default: |
| 300 | return -EINVAL; |
| 301 | } |
| 302 | return 0; |
| 303 | } |
| 304 | |
| 305 | /* Verify a "struct flock" and copy it to a "struct file_lock" as a POSIX |
| 306 | * style lock. |
| 307 | */ |
| 308 | static int flock_to_posix_lock(struct file *filp, struct file_lock *fl, |
| 309 | struct flock *l) |
| 310 | { |
| 311 | off_t start, end; |
| 312 | |
| 313 | switch (l->l_whence) { |
Josef 'Jeff' Sipek | f5579f8 | 2006-09-30 23:27:35 -0700 | [diff] [blame] | 314 | case SEEK_SET: |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 315 | start = 0; |
| 316 | break; |
Josef 'Jeff' Sipek | f5579f8 | 2006-09-30 23:27:35 -0700 | [diff] [blame] | 317 | case SEEK_CUR: |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 318 | start = filp->f_pos; |
| 319 | break; |
Josef 'Jeff' Sipek | f5579f8 | 2006-09-30 23:27:35 -0700 | [diff] [blame] | 320 | case SEEK_END: |
Josef "Jeff" Sipek | 0f7fc9e | 2006-12-08 02:36:35 -0800 | [diff] [blame] | 321 | start = i_size_read(filp->f_path.dentry->d_inode); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 322 | break; |
| 323 | default: |
| 324 | return -EINVAL; |
| 325 | } |
| 326 | |
| 327 | /* POSIX-1996 leaves the case l->l_len < 0 undefined; |
| 328 | POSIX-2001 defines it. */ |
| 329 | start += l->l_start; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 330 | if (start < 0) |
| 331 | return -EINVAL; |
Trond Myklebust | 4c780a4 | 2005-10-18 14:20:21 -0700 | [diff] [blame] | 332 | fl->fl_end = OFFSET_MAX; |
| 333 | if (l->l_len > 0) { |
| 334 | end = start + l->l_len - 1; |
| 335 | fl->fl_end = end; |
| 336 | } else if (l->l_len < 0) { |
| 337 | end = start - 1; |
| 338 | fl->fl_end = end; |
| 339 | start += l->l_len; |
| 340 | if (start < 0) |
| 341 | return -EINVAL; |
| 342 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 343 | fl->fl_start = start; /* we record the absolute position */ |
Trond Myklebust | 4c780a4 | 2005-10-18 14:20:21 -0700 | [diff] [blame] | 344 | if (fl->fl_end < fl->fl_start) |
| 345 | return -EOVERFLOW; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 346 | |
| 347 | fl->fl_owner = current->files; |
| 348 | fl->fl_pid = current->tgid; |
| 349 | fl->fl_file = filp; |
| 350 | fl->fl_flags = FL_POSIX; |
| 351 | fl->fl_ops = NULL; |
| 352 | fl->fl_lmops = NULL; |
| 353 | |
| 354 | return assign_type(fl, l->l_type); |
| 355 | } |
| 356 | |
| 357 | #if BITS_PER_LONG == 32 |
| 358 | static int flock64_to_posix_lock(struct file *filp, struct file_lock *fl, |
| 359 | struct flock64 *l) |
| 360 | { |
| 361 | loff_t start; |
| 362 | |
| 363 | switch (l->l_whence) { |
Josef 'Jeff' Sipek | f5579f8 | 2006-09-30 23:27:35 -0700 | [diff] [blame] | 364 | case SEEK_SET: |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 365 | start = 0; |
| 366 | break; |
Josef 'Jeff' Sipek | f5579f8 | 2006-09-30 23:27:35 -0700 | [diff] [blame] | 367 | case SEEK_CUR: |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 368 | start = filp->f_pos; |
| 369 | break; |
Josef 'Jeff' Sipek | f5579f8 | 2006-09-30 23:27:35 -0700 | [diff] [blame] | 370 | case SEEK_END: |
Josef "Jeff" Sipek | 0f7fc9e | 2006-12-08 02:36:35 -0800 | [diff] [blame] | 371 | start = i_size_read(filp->f_path.dentry->d_inode); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 372 | break; |
| 373 | default: |
| 374 | return -EINVAL; |
| 375 | } |
| 376 | |
Trond Myklebust | 4c780a4 | 2005-10-18 14:20:21 -0700 | [diff] [blame] | 377 | start += l->l_start; |
| 378 | if (start < 0) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 379 | return -EINVAL; |
Trond Myklebust | 4c780a4 | 2005-10-18 14:20:21 -0700 | [diff] [blame] | 380 | fl->fl_end = OFFSET_MAX; |
| 381 | if (l->l_len > 0) { |
| 382 | fl->fl_end = start + l->l_len - 1; |
| 383 | } else if (l->l_len < 0) { |
| 384 | fl->fl_end = start - 1; |
| 385 | start += l->l_len; |
| 386 | if (start < 0) |
| 387 | return -EINVAL; |
| 388 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 389 | fl->fl_start = start; /* we record the absolute position */ |
Trond Myklebust | 4c780a4 | 2005-10-18 14:20:21 -0700 | [diff] [blame] | 390 | if (fl->fl_end < fl->fl_start) |
| 391 | return -EOVERFLOW; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 392 | |
| 393 | fl->fl_owner = current->files; |
| 394 | fl->fl_pid = current->tgid; |
| 395 | fl->fl_file = filp; |
| 396 | fl->fl_flags = FL_POSIX; |
| 397 | fl->fl_ops = NULL; |
| 398 | fl->fl_lmops = NULL; |
| 399 | |
| 400 | switch (l->l_type) { |
| 401 | case F_RDLCK: |
| 402 | case F_WRLCK: |
| 403 | case F_UNLCK: |
| 404 | fl->fl_type = l->l_type; |
| 405 | break; |
| 406 | default: |
| 407 | return -EINVAL; |
| 408 | } |
| 409 | |
| 410 | return (0); |
| 411 | } |
| 412 | #endif |
| 413 | |
| 414 | /* default lease lock manager operations */ |
| 415 | static void lease_break_callback(struct file_lock *fl) |
| 416 | { |
| 417 | kill_fasync(&fl->fl_fasync, SIGIO, POLL_MSG); |
| 418 | } |
| 419 | |
| 420 | static void lease_release_private_callback(struct file_lock *fl) |
| 421 | { |
| 422 | if (!fl->fl_file) |
| 423 | return; |
| 424 | |
| 425 | f_delown(fl->fl_file); |
| 426 | fl->fl_file->f_owner.signum = 0; |
| 427 | } |
| 428 | |
Adrian Bunk | 75c96f8 | 2005-05-05 16:16:09 -0700 | [diff] [blame] | 429 | static int lease_mylease_callback(struct file_lock *fl, struct file_lock *try) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 430 | { |
| 431 | return fl->fl_file == try->fl_file; |
| 432 | } |
| 433 | |
Adrian Bunk | 75c96f8 | 2005-05-05 16:16:09 -0700 | [diff] [blame] | 434 | static struct lock_manager_operations lease_manager_ops = { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 435 | .fl_break = lease_break_callback, |
| 436 | .fl_release_private = lease_release_private_callback, |
| 437 | .fl_mylease = lease_mylease_callback, |
| 438 | .fl_change = lease_modify, |
| 439 | }; |
| 440 | |
| 441 | /* |
| 442 | * Initialize a lease, use the default lock manager operations |
| 443 | */ |
| 444 | static int lease_init(struct file *filp, int type, struct file_lock *fl) |
| 445 | { |
Trond Myklebust | 75dff55 | 2006-05-07 23:02:42 -0400 | [diff] [blame] | 446 | if (assign_type(fl, type) != 0) |
| 447 | return -EINVAL; |
| 448 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 449 | fl->fl_owner = current->files; |
| 450 | fl->fl_pid = current->tgid; |
| 451 | |
| 452 | fl->fl_file = filp; |
| 453 | fl->fl_flags = FL_LEASE; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 454 | fl->fl_start = 0; |
| 455 | fl->fl_end = OFFSET_MAX; |
| 456 | fl->fl_ops = NULL; |
| 457 | fl->fl_lmops = &lease_manager_ops; |
| 458 | return 0; |
| 459 | } |
| 460 | |
| 461 | /* Allocate a file_lock initialised to this type of lease */ |
J. Bruce Fields | e32b8ee | 2007-03-01 14:34:35 -0500 | [diff] [blame] | 462 | static struct file_lock *lease_alloc(struct file *filp, int type) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 463 | { |
| 464 | struct file_lock *fl = locks_alloc_lock(); |
Trond Myklebust | 75dff55 | 2006-05-07 23:02:42 -0400 | [diff] [blame] | 465 | int error = -ENOMEM; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 466 | |
| 467 | if (fl == NULL) |
J. Bruce Fields | e32b8ee | 2007-03-01 14:34:35 -0500 | [diff] [blame] | 468 | return ERR_PTR(error); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 469 | |
| 470 | error = lease_init(filp, type, fl); |
Trond Myklebust | 75dff55 | 2006-05-07 23:02:42 -0400 | [diff] [blame] | 471 | if (error) { |
| 472 | locks_free_lock(fl); |
J. Bruce Fields | e32b8ee | 2007-03-01 14:34:35 -0500 | [diff] [blame] | 473 | return ERR_PTR(error); |
Trond Myklebust | 75dff55 | 2006-05-07 23:02:42 -0400 | [diff] [blame] | 474 | } |
J. Bruce Fields | e32b8ee | 2007-03-01 14:34:35 -0500 | [diff] [blame] | 475 | return fl; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 476 | } |
| 477 | |
| 478 | /* Check if two locks overlap each other. |
| 479 | */ |
| 480 | static inline int locks_overlap(struct file_lock *fl1, struct file_lock *fl2) |
| 481 | { |
| 482 | return ((fl1->fl_end >= fl2->fl_start) && |
| 483 | (fl2->fl_end >= fl1->fl_start)); |
| 484 | } |
| 485 | |
| 486 | /* |
| 487 | * Check whether two locks have the same owner. |
| 488 | */ |
Matt Mackall | 33443c4 | 2006-01-08 01:05:22 -0800 | [diff] [blame] | 489 | static int posix_same_owner(struct file_lock *fl1, struct file_lock *fl2) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 490 | { |
| 491 | if (fl1->fl_lmops && fl1->fl_lmops->fl_compare_owner) |
| 492 | return fl2->fl_lmops == fl1->fl_lmops && |
| 493 | fl1->fl_lmops->fl_compare_owner(fl1, fl2); |
| 494 | return fl1->fl_owner == fl2->fl_owner; |
| 495 | } |
| 496 | |
| 497 | /* Remove waiter from blocker's block list. |
| 498 | * When blocker ends up pointing to itself then the list is empty. |
| 499 | */ |
Matt Mackall | 33443c4 | 2006-01-08 01:05:22 -0800 | [diff] [blame] | 500 | static void __locks_delete_block(struct file_lock *waiter) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 501 | { |
| 502 | list_del_init(&waiter->fl_block); |
| 503 | list_del_init(&waiter->fl_link); |
| 504 | waiter->fl_next = NULL; |
| 505 | } |
| 506 | |
| 507 | /* |
| 508 | */ |
| 509 | static void locks_delete_block(struct file_lock *waiter) |
| 510 | { |
| 511 | lock_kernel(); |
| 512 | __locks_delete_block(waiter); |
| 513 | unlock_kernel(); |
| 514 | } |
| 515 | |
| 516 | /* Insert waiter into blocker's block list. |
| 517 | * We use a circular list so that processes can be easily woken up in |
| 518 | * the order they blocked. The documentation doesn't require this but |
| 519 | * it seems like the reasonable thing to do. |
| 520 | */ |
| 521 | static void locks_insert_block(struct file_lock *blocker, |
| 522 | struct file_lock *waiter) |
| 523 | { |
J. Bruce Fields | 6dc0fe8 | 2006-03-26 01:37:24 -0800 | [diff] [blame] | 524 | BUG_ON(!list_empty(&waiter->fl_block)); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 525 | list_add_tail(&waiter->fl_block, &blocker->fl_block); |
| 526 | waiter->fl_next = blocker; |
| 527 | if (IS_POSIX(blocker)) |
| 528 | list_add(&waiter->fl_link, &blocked_list); |
| 529 | } |
| 530 | |
| 531 | /* Wake up processes blocked waiting for blocker. |
| 532 | * If told to wait then schedule the processes until the block list |
| 533 | * is empty, otherwise empty the block list ourselves. |
| 534 | */ |
| 535 | static void locks_wake_up_blocks(struct file_lock *blocker) |
| 536 | { |
| 537 | while (!list_empty(&blocker->fl_block)) { |
Pavel Emelyanov | f0c1cd0 | 2007-09-19 16:44:07 +0400 | [diff] [blame] | 538 | struct file_lock *waiter; |
| 539 | |
| 540 | waiter = list_first_entry(&blocker->fl_block, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 541 | struct file_lock, fl_block); |
| 542 | __locks_delete_block(waiter); |
| 543 | if (waiter->fl_lmops && waiter->fl_lmops->fl_notify) |
| 544 | waiter->fl_lmops->fl_notify(waiter); |
| 545 | else |
| 546 | wake_up(&waiter->fl_wait); |
| 547 | } |
| 548 | } |
| 549 | |
| 550 | /* Insert file lock fl into an inode's lock list at the position indicated |
| 551 | * by pos. At the same time add the lock to the global file lock list. |
| 552 | */ |
| 553 | static void locks_insert_lock(struct file_lock **pos, struct file_lock *fl) |
| 554 | { |
| 555 | list_add(&fl->fl_link, &file_lock_list); |
| 556 | |
Vitaliy Gusev | ab1f161 | 2008-01-17 00:07:08 +0000 | [diff] [blame] | 557 | fl->fl_nspid = get_pid(task_tgid(current)); |
| 558 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 559 | /* insert into file's list */ |
| 560 | fl->fl_next = *pos; |
| 561 | *pos = fl; |
| 562 | |
| 563 | if (fl->fl_ops && fl->fl_ops->fl_insert) |
| 564 | fl->fl_ops->fl_insert(fl); |
| 565 | } |
| 566 | |
| 567 | /* |
| 568 | * Delete a lock and then free it. |
| 569 | * Wake up processes that are blocked waiting for this lock, |
| 570 | * notify the FS that the lock has been cleared and |
| 571 | * finally free the lock. |
| 572 | */ |
| 573 | static void locks_delete_lock(struct file_lock **thisfl_p) |
| 574 | { |
| 575 | struct file_lock *fl = *thisfl_p; |
| 576 | |
| 577 | *thisfl_p = fl->fl_next; |
| 578 | fl->fl_next = NULL; |
| 579 | list_del_init(&fl->fl_link); |
| 580 | |
| 581 | fasync_helper(0, fl->fl_file, 0, &fl->fl_fasync); |
| 582 | if (fl->fl_fasync != NULL) { |
| 583 | printk(KERN_ERR "locks_delete_lock: fasync == %p\n", fl->fl_fasync); |
| 584 | fl->fl_fasync = NULL; |
| 585 | } |
| 586 | |
| 587 | if (fl->fl_ops && fl->fl_ops->fl_remove) |
| 588 | fl->fl_ops->fl_remove(fl); |
| 589 | |
Vitaliy Gusev | ab1f161 | 2008-01-17 00:07:08 +0000 | [diff] [blame] | 590 | if (fl->fl_nspid) { |
| 591 | put_pid(fl->fl_nspid); |
| 592 | fl->fl_nspid = NULL; |
| 593 | } |
| 594 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 595 | locks_wake_up_blocks(fl); |
| 596 | locks_free_lock(fl); |
| 597 | } |
| 598 | |
| 599 | /* Determine if lock sys_fl blocks lock caller_fl. Common functionality |
| 600 | * checks for shared/exclusive status of overlapping locks. |
| 601 | */ |
| 602 | static int locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) |
| 603 | { |
| 604 | if (sys_fl->fl_type == F_WRLCK) |
| 605 | return 1; |
| 606 | if (caller_fl->fl_type == F_WRLCK) |
| 607 | return 1; |
| 608 | return 0; |
| 609 | } |
| 610 | |
| 611 | /* Determine if lock sys_fl blocks lock caller_fl. POSIX specific |
| 612 | * checking before calling the locks_conflict(). |
| 613 | */ |
| 614 | static int posix_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) |
| 615 | { |
| 616 | /* POSIX locks owned by the same process do not conflict with |
| 617 | * each other. |
| 618 | */ |
| 619 | if (!IS_POSIX(sys_fl) || posix_same_owner(caller_fl, sys_fl)) |
| 620 | return (0); |
| 621 | |
| 622 | /* Check whether they overlap */ |
| 623 | if (!locks_overlap(caller_fl, sys_fl)) |
| 624 | return 0; |
| 625 | |
| 626 | return (locks_conflict(caller_fl, sys_fl)); |
| 627 | } |
| 628 | |
| 629 | /* Determine if lock sys_fl blocks lock caller_fl. FLOCK specific |
| 630 | * checking before calling the locks_conflict(). |
| 631 | */ |
| 632 | static int flock_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) |
| 633 | { |
| 634 | /* FLOCK locks referring to the same filp do not conflict with |
| 635 | * each other. |
| 636 | */ |
| 637 | if (!IS_FLOCK(sys_fl) || (caller_fl->fl_file == sys_fl->fl_file)) |
| 638 | return (0); |
| 639 | if ((caller_fl->fl_type & LOCK_MAND) || (sys_fl->fl_type & LOCK_MAND)) |
| 640 | return 0; |
| 641 | |
| 642 | return (locks_conflict(caller_fl, sys_fl)); |
| 643 | } |
| 644 | |
J. Bruce Fields | 6d34ac1 | 2007-05-11 16:09:32 -0400 | [diff] [blame] | 645 | void |
Marc Eshel | 9d6a8c5 | 2007-02-21 00:55:18 -0500 | [diff] [blame] | 646 | posix_test_lock(struct file *filp, struct file_lock *fl) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 647 | { |
| 648 | struct file_lock *cfl; |
| 649 | |
| 650 | lock_kernel(); |
Josef "Jeff" Sipek | 0f7fc9e | 2006-12-08 02:36:35 -0800 | [diff] [blame] | 651 | for (cfl = filp->f_path.dentry->d_inode->i_flock; cfl; cfl = cfl->fl_next) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 652 | if (!IS_POSIX(cfl)) |
| 653 | continue; |
J. Bruce Fields | b842e24 | 2007-05-10 19:02:07 -0400 | [diff] [blame] | 654 | if (posix_locks_conflict(fl, cfl)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 655 | break; |
| 656 | } |
Vitaliy Gusev | ab1f161 | 2008-01-17 00:07:08 +0000 | [diff] [blame] | 657 | if (cfl) { |
Marc Eshel | 9d6a8c5 | 2007-02-21 00:55:18 -0500 | [diff] [blame] | 658 | __locks_copy_lock(fl, cfl); |
Vitaliy Gusev | ab1f161 | 2008-01-17 00:07:08 +0000 | [diff] [blame] | 659 | if (cfl->fl_nspid) |
Pavel Emelyanov | 6c5f3e7 | 2008-02-08 04:19:20 -0800 | [diff] [blame] | 660 | fl->fl_pid = pid_vnr(cfl->fl_nspid); |
Vitaliy Gusev | ab1f161 | 2008-01-17 00:07:08 +0000 | [diff] [blame] | 661 | } else |
J. Bruce Fields | 129a84d | 2007-05-10 18:38:43 -0400 | [diff] [blame] | 662 | fl->fl_type = F_UNLCK; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 663 | unlock_kernel(); |
J. Bruce Fields | 6d34ac1 | 2007-05-11 16:09:32 -0400 | [diff] [blame] | 664 | return; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 665 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 666 | EXPORT_SYMBOL(posix_test_lock); |
| 667 | |
J. Bruce Fields | b533184 | 2007-10-26 18:05:40 -0400 | [diff] [blame] | 668 | /* |
| 669 | * Deadlock detection: |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 670 | * |
J. Bruce Fields | b533184 | 2007-10-26 18:05:40 -0400 | [diff] [blame] | 671 | * We attempt to detect deadlocks that are due purely to posix file |
| 672 | * locks. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 673 | * |
J. Bruce Fields | b533184 | 2007-10-26 18:05:40 -0400 | [diff] [blame] | 674 | * We assume that a task can be waiting for at most one lock at a time. |
| 675 | * So for any acquired lock, the process holding that lock may be |
| 676 | * waiting on at most one other lock. That lock in turns may be held by |
| 677 | * someone waiting for at most one other lock. Given a requested lock |
| 678 | * caller_fl which is about to wait for a conflicting lock block_fl, we |
| 679 | * follow this chain of waiters to ensure we are not about to create a |
| 680 | * cycle. |
J. Bruce Fields | 97855b4 | 2007-10-30 11:20:02 -0400 | [diff] [blame] | 681 | * |
J. Bruce Fields | b533184 | 2007-10-26 18:05:40 -0400 | [diff] [blame] | 682 | * Since we do this before we ever put a process to sleep on a lock, we |
| 683 | * are ensured that there is never a cycle; that is what guarantees that |
| 684 | * the while() loop in posix_locks_deadlock() eventually completes. |
| 685 | * |
| 686 | * Note: the above assumption may not be true when handling lock |
| 687 | * requests from a broken NFS client. It may also fail in the presence |
| 688 | * of tasks (such as posix threads) sharing the same open file table. |
| 689 | * |
| 690 | * To handle those cases, we just bail out after a few iterations. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 691 | */ |
J. Bruce Fields | 97855b4 | 2007-10-30 11:20:02 -0400 | [diff] [blame] | 692 | |
| 693 | #define MAX_DEADLK_ITERATIONS 10 |
| 694 | |
J. Bruce Fields | b533184 | 2007-10-26 18:05:40 -0400 | [diff] [blame] | 695 | /* Find a lock that the owner of the given block_fl is blocking on. */ |
| 696 | static struct file_lock *what_owner_is_waiting_for(struct file_lock *block_fl) |
| 697 | { |
| 698 | struct file_lock *fl; |
| 699 | |
| 700 | list_for_each_entry(fl, &blocked_list, fl_link) { |
| 701 | if (posix_same_owner(fl, block_fl)) |
| 702 | return fl->fl_next; |
| 703 | } |
| 704 | return NULL; |
| 705 | } |
| 706 | |
Adrian Bunk | b0904e1 | 2006-06-23 02:05:13 -0700 | [diff] [blame] | 707 | static int posix_locks_deadlock(struct file_lock *caller_fl, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 708 | struct file_lock *block_fl) |
| 709 | { |
J. Bruce Fields | 97855b4 | 2007-10-30 11:20:02 -0400 | [diff] [blame] | 710 | int i = 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 711 | |
J. Bruce Fields | b533184 | 2007-10-26 18:05:40 -0400 | [diff] [blame] | 712 | while ((block_fl = what_owner_is_waiting_for(block_fl))) { |
| 713 | if (i++ > MAX_DEADLK_ITERATIONS) |
| 714 | return 0; |
| 715 | if (posix_same_owner(caller_fl, block_fl)) |
| 716 | return 1; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 717 | } |
| 718 | return 0; |
| 719 | } |
| 720 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 721 | /* Try to create a FLOCK lock on filp. We always insert new FLOCK locks |
J. Bruce Fields | 02888f4 | 2007-09-12 15:45:07 -0400 | [diff] [blame] | 722 | * after any leases, but before any posix locks. |
Trond Myklebust | f475ae9 | 2006-06-29 16:38:32 -0400 | [diff] [blame] | 723 | * |
| 724 | * Note that if called with an FL_EXISTS argument, the caller may determine |
| 725 | * whether or not a lock was successfully freed by testing the return |
| 726 | * value for -ENOENT. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 727 | */ |
Trond Myklebust | 993dfa8 | 2006-03-31 02:30:55 -0800 | [diff] [blame] | 728 | static int flock_lock_file(struct file *filp, struct file_lock *request) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 729 | { |
Trond Myklebust | 993dfa8 | 2006-03-31 02:30:55 -0800 | [diff] [blame] | 730 | struct file_lock *new_fl = NULL; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 731 | struct file_lock **before; |
Josef "Jeff" Sipek | 0f7fc9e | 2006-12-08 02:36:35 -0800 | [diff] [blame] | 732 | struct inode * inode = filp->f_path.dentry->d_inode; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 733 | int error = 0; |
| 734 | int found = 0; |
| 735 | |
| 736 | lock_kernel(); |
Trond Myklebust | f07f18d | 2006-06-29 16:38:37 -0400 | [diff] [blame] | 737 | if (request->fl_flags & FL_ACCESS) |
| 738 | goto find_conflict; |
Pavel Emelyanov | 84d535a | 2007-09-11 16:38:13 +0400 | [diff] [blame] | 739 | |
| 740 | if (request->fl_type != F_UNLCK) { |
| 741 | error = -ENOMEM; |
| 742 | new_fl = locks_alloc_lock(); |
| 743 | if (new_fl == NULL) |
| 744 | goto out; |
| 745 | error = 0; |
| 746 | } |
| 747 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 748 | for_each_lock(inode, before) { |
| 749 | struct file_lock *fl = *before; |
| 750 | if (IS_POSIX(fl)) |
| 751 | break; |
| 752 | if (IS_LEASE(fl)) |
| 753 | continue; |
| 754 | if (filp != fl->fl_file) |
| 755 | continue; |
Trond Myklebust | 993dfa8 | 2006-03-31 02:30:55 -0800 | [diff] [blame] | 756 | if (request->fl_type == fl->fl_type) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 757 | goto out; |
| 758 | found = 1; |
| 759 | locks_delete_lock(before); |
| 760 | break; |
| 761 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 762 | |
Trond Myklebust | f475ae9 | 2006-06-29 16:38:32 -0400 | [diff] [blame] | 763 | if (request->fl_type == F_UNLCK) { |
| 764 | if ((request->fl_flags & FL_EXISTS) && !found) |
| 765 | error = -ENOENT; |
Trond Myklebust | 993dfa8 | 2006-03-31 02:30:55 -0800 | [diff] [blame] | 766 | goto out; |
Trond Myklebust | f475ae9 | 2006-06-29 16:38:32 -0400 | [diff] [blame] | 767 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 768 | |
| 769 | /* |
| 770 | * If a higher-priority process was blocked on the old file lock, |
| 771 | * give it the opportunity to lock the file. |
| 772 | */ |
| 773 | if (found) |
| 774 | cond_resched(); |
| 775 | |
Trond Myklebust | f07f18d | 2006-06-29 16:38:37 -0400 | [diff] [blame] | 776 | find_conflict: |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 777 | for_each_lock(inode, before) { |
| 778 | struct file_lock *fl = *before; |
| 779 | if (IS_POSIX(fl)) |
| 780 | break; |
| 781 | if (IS_LEASE(fl)) |
| 782 | continue; |
Trond Myklebust | 993dfa8 | 2006-03-31 02:30:55 -0800 | [diff] [blame] | 783 | if (!flock_locks_conflict(request, fl)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 784 | continue; |
| 785 | error = -EAGAIN; |
Trond Myklebust | 993dfa8 | 2006-03-31 02:30:55 -0800 | [diff] [blame] | 786 | if (request->fl_flags & FL_SLEEP) |
| 787 | locks_insert_block(fl, request); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 788 | goto out; |
| 789 | } |
Trond Myklebust | f07f18d | 2006-06-29 16:38:37 -0400 | [diff] [blame] | 790 | if (request->fl_flags & FL_ACCESS) |
| 791 | goto out; |
Trond Myklebust | 993dfa8 | 2006-03-31 02:30:55 -0800 | [diff] [blame] | 792 | locks_copy_lock(new_fl, request); |
Pavel Emelyanov | 0e2f6db | 2007-09-11 15:24:01 -0700 | [diff] [blame] | 793 | locks_insert_lock(before, new_fl); |
Trond Myklebust | 993dfa8 | 2006-03-31 02:30:55 -0800 | [diff] [blame] | 794 | new_fl = NULL; |
Kirill Korotaev | 9cedc19 | 2006-06-14 17:59:35 +0400 | [diff] [blame] | 795 | error = 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 796 | |
| 797 | out: |
| 798 | unlock_kernel(); |
Trond Myklebust | 993dfa8 | 2006-03-31 02:30:55 -0800 | [diff] [blame] | 799 | if (new_fl) |
| 800 | locks_free_lock(new_fl); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 801 | return error; |
| 802 | } |
| 803 | |
Marc Eshel | 150b393 | 2007-01-18 16:15:35 -0500 | [diff] [blame] | 804 | static int __posix_lock_file(struct inode *inode, struct file_lock *request, struct file_lock *conflock) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 805 | { |
| 806 | struct file_lock *fl; |
Miklos Szeredi | 39005d0 | 2006-06-23 02:05:10 -0700 | [diff] [blame] | 807 | struct file_lock *new_fl = NULL; |
| 808 | struct file_lock *new_fl2 = NULL; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 809 | struct file_lock *left = NULL; |
| 810 | struct file_lock *right = NULL; |
| 811 | struct file_lock **before; |
| 812 | int error, added = 0; |
| 813 | |
| 814 | /* |
| 815 | * We may need two file_lock structures for this operation, |
| 816 | * so we get them in advance to avoid races. |
Miklos Szeredi | 39005d0 | 2006-06-23 02:05:10 -0700 | [diff] [blame] | 817 | * |
| 818 | * In some cases we can be sure, that no new locks will be needed |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 819 | */ |
Miklos Szeredi | 39005d0 | 2006-06-23 02:05:10 -0700 | [diff] [blame] | 820 | if (!(request->fl_flags & FL_ACCESS) && |
| 821 | (request->fl_type != F_UNLCK || |
| 822 | request->fl_start != 0 || request->fl_end != OFFSET_MAX)) { |
| 823 | new_fl = locks_alloc_lock(); |
| 824 | new_fl2 = locks_alloc_lock(); |
| 825 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 826 | |
| 827 | lock_kernel(); |
| 828 | if (request->fl_type != F_UNLCK) { |
| 829 | for_each_lock(inode, before) { |
J. Bruce Fields | 526985b | 2006-11-14 16:54:36 -0500 | [diff] [blame] | 830 | fl = *before; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 831 | if (!IS_POSIX(fl)) |
| 832 | continue; |
| 833 | if (!posix_locks_conflict(request, fl)) |
| 834 | continue; |
Andy Adamson | 5842add | 2006-03-26 01:37:26 -0800 | [diff] [blame] | 835 | if (conflock) |
J. Bruce Fields | 1a747ee | 2008-04-24 10:08:22 -0400 | [diff] [blame] | 836 | __locks_copy_lock(conflock, fl); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 837 | error = -EAGAIN; |
| 838 | if (!(request->fl_flags & FL_SLEEP)) |
| 839 | goto out; |
| 840 | error = -EDEADLK; |
| 841 | if (posix_locks_deadlock(request, fl)) |
| 842 | goto out; |
| 843 | error = -EAGAIN; |
| 844 | locks_insert_block(fl, request); |
| 845 | goto out; |
| 846 | } |
| 847 | } |
| 848 | |
| 849 | /* If we're just looking for a conflict, we're done. */ |
| 850 | error = 0; |
| 851 | if (request->fl_flags & FL_ACCESS) |
| 852 | goto out; |
| 853 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 854 | /* |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 855 | * Find the first old lock with the same owner as the new lock. |
| 856 | */ |
| 857 | |
| 858 | before = &inode->i_flock; |
| 859 | |
| 860 | /* First skip locks owned by other processes. */ |
| 861 | while ((fl = *before) && (!IS_POSIX(fl) || |
| 862 | !posix_same_owner(request, fl))) { |
| 863 | before = &fl->fl_next; |
| 864 | } |
| 865 | |
| 866 | /* Process locks with this owner. */ |
| 867 | while ((fl = *before) && posix_same_owner(request, fl)) { |
| 868 | /* Detect adjacent or overlapping regions (if same lock type) |
| 869 | */ |
| 870 | if (request->fl_type == fl->fl_type) { |
Olaf Kirch | 449231d | 2005-08-25 16:25:35 -0700 | [diff] [blame] | 871 | /* In all comparisons of start vs end, use |
| 872 | * "start - 1" rather than "end + 1". If end |
| 873 | * is OFFSET_MAX, end + 1 will become negative. |
| 874 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 875 | if (fl->fl_end < request->fl_start - 1) |
| 876 | goto next_lock; |
| 877 | /* If the next lock in the list has entirely bigger |
| 878 | * addresses than the new one, insert the lock here. |
| 879 | */ |
Olaf Kirch | 449231d | 2005-08-25 16:25:35 -0700 | [diff] [blame] | 880 | if (fl->fl_start - 1 > request->fl_end) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 881 | break; |
| 882 | |
| 883 | /* If we come here, the new and old lock are of the |
| 884 | * same type and adjacent or overlapping. Make one |
| 885 | * lock yielding from the lower start address of both |
| 886 | * locks to the higher end address. |
| 887 | */ |
| 888 | if (fl->fl_start > request->fl_start) |
| 889 | fl->fl_start = request->fl_start; |
| 890 | else |
| 891 | request->fl_start = fl->fl_start; |
| 892 | if (fl->fl_end < request->fl_end) |
| 893 | fl->fl_end = request->fl_end; |
| 894 | else |
| 895 | request->fl_end = fl->fl_end; |
| 896 | if (added) { |
| 897 | locks_delete_lock(before); |
| 898 | continue; |
| 899 | } |
| 900 | request = fl; |
| 901 | added = 1; |
| 902 | } |
| 903 | else { |
| 904 | /* Processing for different lock types is a bit |
| 905 | * more complex. |
| 906 | */ |
| 907 | if (fl->fl_end < request->fl_start) |
| 908 | goto next_lock; |
| 909 | if (fl->fl_start > request->fl_end) |
| 910 | break; |
| 911 | if (request->fl_type == F_UNLCK) |
| 912 | added = 1; |
| 913 | if (fl->fl_start < request->fl_start) |
| 914 | left = fl; |
| 915 | /* If the next lock in the list has a higher end |
| 916 | * address than the new one, insert the new one here. |
| 917 | */ |
| 918 | if (fl->fl_end > request->fl_end) { |
| 919 | right = fl; |
| 920 | break; |
| 921 | } |
| 922 | if (fl->fl_start >= request->fl_start) { |
| 923 | /* The new lock completely replaces an old |
| 924 | * one (This may happen several times). |
| 925 | */ |
| 926 | if (added) { |
| 927 | locks_delete_lock(before); |
| 928 | continue; |
| 929 | } |
| 930 | /* Replace the old lock with the new one. |
| 931 | * Wake up anybody waiting for the old one, |
| 932 | * as the change in lock type might satisfy |
| 933 | * their needs. |
| 934 | */ |
| 935 | locks_wake_up_blocks(fl); |
| 936 | fl->fl_start = request->fl_start; |
| 937 | fl->fl_end = request->fl_end; |
| 938 | fl->fl_type = request->fl_type; |
Trond Myklebust | 47831f3 | 2006-03-20 13:44:05 -0500 | [diff] [blame] | 939 | locks_release_private(fl); |
| 940 | locks_copy_private(fl, request); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 941 | request = fl; |
| 942 | added = 1; |
| 943 | } |
| 944 | } |
| 945 | /* Go on to next lock. |
| 946 | */ |
| 947 | next_lock: |
| 948 | before = &fl->fl_next; |
| 949 | } |
| 950 | |
Miklos Szeredi | 0d9a490 | 2006-06-23 02:05:09 -0700 | [diff] [blame] | 951 | /* |
| 952 | * The above code only modifies existing locks in case of |
| 953 | * merging or replacing. If new lock(s) need to be inserted |
| 954 | * all modifications are done bellow this, so it's safe yet to |
| 955 | * bail out. |
| 956 | */ |
| 957 | error = -ENOLCK; /* "no luck" */ |
| 958 | if (right && left == right && !new_fl2) |
| 959 | goto out; |
| 960 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 961 | error = 0; |
| 962 | if (!added) { |
Trond Myklebust | f475ae9 | 2006-06-29 16:38:32 -0400 | [diff] [blame] | 963 | if (request->fl_type == F_UNLCK) { |
| 964 | if (request->fl_flags & FL_EXISTS) |
| 965 | error = -ENOENT; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 966 | goto out; |
Trond Myklebust | f475ae9 | 2006-06-29 16:38:32 -0400 | [diff] [blame] | 967 | } |
Miklos Szeredi | 0d9a490 | 2006-06-23 02:05:09 -0700 | [diff] [blame] | 968 | |
| 969 | if (!new_fl) { |
| 970 | error = -ENOLCK; |
| 971 | goto out; |
| 972 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 973 | locks_copy_lock(new_fl, request); |
| 974 | locks_insert_lock(before, new_fl); |
| 975 | new_fl = NULL; |
| 976 | } |
| 977 | if (right) { |
| 978 | if (left == right) { |
| 979 | /* The new lock breaks the old one in two pieces, |
| 980 | * so we have to use the second new lock. |
| 981 | */ |
| 982 | left = new_fl2; |
| 983 | new_fl2 = NULL; |
| 984 | locks_copy_lock(left, right); |
| 985 | locks_insert_lock(before, left); |
| 986 | } |
| 987 | right->fl_start = request->fl_end + 1; |
| 988 | locks_wake_up_blocks(right); |
| 989 | } |
| 990 | if (left) { |
| 991 | left->fl_end = request->fl_start - 1; |
| 992 | locks_wake_up_blocks(left); |
| 993 | } |
| 994 | out: |
| 995 | unlock_kernel(); |
| 996 | /* |
| 997 | * Free any unused locks. |
| 998 | */ |
| 999 | if (new_fl) |
| 1000 | locks_free_lock(new_fl); |
| 1001 | if (new_fl2) |
| 1002 | locks_free_lock(new_fl2); |
| 1003 | return error; |
| 1004 | } |
| 1005 | |
| 1006 | /** |
| 1007 | * posix_lock_file - Apply a POSIX-style lock to a file |
| 1008 | * @filp: The file to apply the lock to |
| 1009 | * @fl: The lock to be applied |
Marc Eshel | 150b393 | 2007-01-18 16:15:35 -0500 | [diff] [blame] | 1010 | * @conflock: Place to return a copy of the conflicting lock, if found. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1011 | * |
| 1012 | * Add a POSIX style lock to a file. |
| 1013 | * We merge adjacent & overlapping locks whenever possible. |
| 1014 | * POSIX locks are sorted by owner task, then by starting address |
Trond Myklebust | f475ae9 | 2006-06-29 16:38:32 -0400 | [diff] [blame] | 1015 | * |
| 1016 | * Note that if called with an FL_EXISTS argument, the caller may determine |
| 1017 | * whether or not a lock was successfully freed by testing the return |
| 1018 | * value for -ENOENT. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1019 | */ |
Marc Eshel | 150b393 | 2007-01-18 16:15:35 -0500 | [diff] [blame] | 1020 | int posix_lock_file(struct file *filp, struct file_lock *fl, |
Andy Adamson | 5842add | 2006-03-26 01:37:26 -0800 | [diff] [blame] | 1021 | struct file_lock *conflock) |
| 1022 | { |
Marc Eshel | 150b393 | 2007-01-18 16:15:35 -0500 | [diff] [blame] | 1023 | return __posix_lock_file(filp->f_path.dentry->d_inode, fl, conflock); |
Andy Adamson | 5842add | 2006-03-26 01:37:26 -0800 | [diff] [blame] | 1024 | } |
Marc Eshel | 150b393 | 2007-01-18 16:15:35 -0500 | [diff] [blame] | 1025 | EXPORT_SYMBOL(posix_lock_file); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1026 | |
| 1027 | /** |
| 1028 | * posix_lock_file_wait - Apply a POSIX-style lock to a file |
| 1029 | * @filp: The file to apply the lock to |
| 1030 | * @fl: The lock to be applied |
| 1031 | * |
| 1032 | * Add a POSIX style lock to a file. |
| 1033 | * We merge adjacent & overlapping locks whenever possible. |
| 1034 | * POSIX locks are sorted by owner task, then by starting address |
| 1035 | */ |
| 1036 | int posix_lock_file_wait(struct file *filp, struct file_lock *fl) |
| 1037 | { |
| 1038 | int error; |
| 1039 | might_sleep (); |
| 1040 | for (;;) { |
Marc Eshel | 150b393 | 2007-01-18 16:15:35 -0500 | [diff] [blame] | 1041 | error = posix_lock_file(filp, fl, NULL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1042 | if ((error != -EAGAIN) || !(fl->fl_flags & FL_SLEEP)) |
| 1043 | break; |
| 1044 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_next); |
| 1045 | if (!error) |
| 1046 | continue; |
| 1047 | |
| 1048 | locks_delete_block(fl); |
| 1049 | break; |
| 1050 | } |
| 1051 | return error; |
| 1052 | } |
| 1053 | EXPORT_SYMBOL(posix_lock_file_wait); |
| 1054 | |
| 1055 | /** |
| 1056 | * locks_mandatory_locked - Check for an active lock |
| 1057 | * @inode: the file to check |
| 1058 | * |
| 1059 | * Searches the inode's list of locks to find any POSIX locks which conflict. |
| 1060 | * This function is called from locks_verify_locked() only. |
| 1061 | */ |
| 1062 | int locks_mandatory_locked(struct inode *inode) |
| 1063 | { |
| 1064 | fl_owner_t owner = current->files; |
| 1065 | struct file_lock *fl; |
| 1066 | |
| 1067 | /* |
| 1068 | * Search the lock list for this inode for any POSIX locks. |
| 1069 | */ |
| 1070 | lock_kernel(); |
| 1071 | for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) { |
| 1072 | if (!IS_POSIX(fl)) |
| 1073 | continue; |
| 1074 | if (fl->fl_owner != owner) |
| 1075 | break; |
| 1076 | } |
| 1077 | unlock_kernel(); |
| 1078 | return fl ? -EAGAIN : 0; |
| 1079 | } |
| 1080 | |
| 1081 | /** |
| 1082 | * locks_mandatory_area - Check for a conflicting lock |
| 1083 | * @read_write: %FLOCK_VERIFY_WRITE for exclusive access, %FLOCK_VERIFY_READ |
| 1084 | * for shared |
| 1085 | * @inode: the file to check |
| 1086 | * @filp: how the file was opened (if it was) |
| 1087 | * @offset: start of area to check |
| 1088 | * @count: length of area to check |
| 1089 | * |
| 1090 | * Searches the inode's list of locks to find any POSIX locks which conflict. |
| 1091 | * This function is called from rw_verify_area() and |
| 1092 | * locks_verify_truncate(). |
| 1093 | */ |
| 1094 | int locks_mandatory_area(int read_write, struct inode *inode, |
| 1095 | struct file *filp, loff_t offset, |
| 1096 | size_t count) |
| 1097 | { |
| 1098 | struct file_lock fl; |
| 1099 | int error; |
| 1100 | |
| 1101 | locks_init_lock(&fl); |
| 1102 | fl.fl_owner = current->files; |
| 1103 | fl.fl_pid = current->tgid; |
| 1104 | fl.fl_file = filp; |
| 1105 | fl.fl_flags = FL_POSIX | FL_ACCESS; |
| 1106 | if (filp && !(filp->f_flags & O_NONBLOCK)) |
| 1107 | fl.fl_flags |= FL_SLEEP; |
| 1108 | fl.fl_type = (read_write == FLOCK_VERIFY_WRITE) ? F_WRLCK : F_RDLCK; |
| 1109 | fl.fl_start = offset; |
| 1110 | fl.fl_end = offset + count - 1; |
| 1111 | |
| 1112 | for (;;) { |
Marc Eshel | 150b393 | 2007-01-18 16:15:35 -0500 | [diff] [blame] | 1113 | error = __posix_lock_file(inode, &fl, NULL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1114 | if (error != -EAGAIN) |
| 1115 | break; |
| 1116 | if (!(fl.fl_flags & FL_SLEEP)) |
| 1117 | break; |
| 1118 | error = wait_event_interruptible(fl.fl_wait, !fl.fl_next); |
| 1119 | if (!error) { |
| 1120 | /* |
| 1121 | * If we've been sleeping someone might have |
| 1122 | * changed the permissions behind our back. |
| 1123 | */ |
Pavel Emelyanov | a16877c | 2007-10-01 14:41:11 -0700 | [diff] [blame] | 1124 | if (__mandatory_lock(inode)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1125 | continue; |
| 1126 | } |
| 1127 | |
| 1128 | locks_delete_block(&fl); |
| 1129 | break; |
| 1130 | } |
| 1131 | |
| 1132 | return error; |
| 1133 | } |
| 1134 | |
| 1135 | EXPORT_SYMBOL(locks_mandatory_area); |
| 1136 | |
| 1137 | /* We already had a lease on this file; just change its type */ |
| 1138 | int lease_modify(struct file_lock **before, int arg) |
| 1139 | { |
| 1140 | struct file_lock *fl = *before; |
| 1141 | int error = assign_type(fl, arg); |
| 1142 | |
| 1143 | if (error) |
| 1144 | return error; |
| 1145 | locks_wake_up_blocks(fl); |
| 1146 | if (arg == F_UNLCK) |
| 1147 | locks_delete_lock(before); |
| 1148 | return 0; |
| 1149 | } |
| 1150 | |
| 1151 | EXPORT_SYMBOL(lease_modify); |
| 1152 | |
| 1153 | static void time_out_leases(struct inode *inode) |
| 1154 | { |
| 1155 | struct file_lock **before; |
| 1156 | struct file_lock *fl; |
| 1157 | |
| 1158 | before = &inode->i_flock; |
| 1159 | while ((fl = *before) && IS_LEASE(fl) && (fl->fl_type & F_INPROGRESS)) { |
| 1160 | if ((fl->fl_break_time == 0) |
| 1161 | || time_before(jiffies, fl->fl_break_time)) { |
| 1162 | before = &fl->fl_next; |
| 1163 | continue; |
| 1164 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1165 | lease_modify(before, fl->fl_type & ~F_INPROGRESS); |
| 1166 | if (fl == *before) /* lease_modify may have freed fl */ |
| 1167 | before = &fl->fl_next; |
| 1168 | } |
| 1169 | } |
| 1170 | |
| 1171 | /** |
| 1172 | * __break_lease - revoke all outstanding leases on file |
| 1173 | * @inode: the inode of the file to return |
| 1174 | * @mode: the open mode (read or write) |
| 1175 | * |
david m. richter | 87250dd | 2007-05-09 16:10:27 -0400 | [diff] [blame] | 1176 | * break_lease (inlined for speed) has checked there already is at least |
| 1177 | * some kind of lock (maybe a lease) on this file. Leases are broken on |
| 1178 | * a call to open() or truncate(). This function can sleep unless you |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1179 | * specified %O_NONBLOCK to your open(). |
| 1180 | */ |
| 1181 | int __break_lease(struct inode *inode, unsigned int mode) |
| 1182 | { |
| 1183 | int error = 0, future; |
| 1184 | struct file_lock *new_fl, *flock; |
| 1185 | struct file_lock *fl; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1186 | unsigned long break_time; |
| 1187 | int i_have_this_lease = 0; |
| 1188 | |
J. Bruce Fields | e32b8ee | 2007-03-01 14:34:35 -0500 | [diff] [blame] | 1189 | new_fl = lease_alloc(NULL, mode & FMODE_WRITE ? F_WRLCK : F_RDLCK); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1190 | |
| 1191 | lock_kernel(); |
| 1192 | |
| 1193 | time_out_leases(inode); |
| 1194 | |
| 1195 | flock = inode->i_flock; |
| 1196 | if ((flock == NULL) || !IS_LEASE(flock)) |
| 1197 | goto out; |
| 1198 | |
| 1199 | for (fl = flock; fl && IS_LEASE(fl); fl = fl->fl_next) |
| 1200 | if (fl->fl_owner == current->files) |
| 1201 | i_have_this_lease = 1; |
| 1202 | |
| 1203 | if (mode & FMODE_WRITE) { |
| 1204 | /* If we want write access, we have to revoke any lease. */ |
| 1205 | future = F_UNLCK | F_INPROGRESS; |
| 1206 | } else if (flock->fl_type & F_INPROGRESS) { |
| 1207 | /* If the lease is already being broken, we just leave it */ |
| 1208 | future = flock->fl_type; |
| 1209 | } else if (flock->fl_type & F_WRLCK) { |
| 1210 | /* Downgrade the exclusive lease to a read-only lease. */ |
| 1211 | future = F_RDLCK | F_INPROGRESS; |
| 1212 | } else { |
| 1213 | /* the existing lease was read-only, so we can read too. */ |
| 1214 | goto out; |
| 1215 | } |
| 1216 | |
J. Bruce Fields | e32b8ee | 2007-03-01 14:34:35 -0500 | [diff] [blame] | 1217 | if (IS_ERR(new_fl) && !i_have_this_lease |
| 1218 | && ((mode & O_NONBLOCK) == 0)) { |
| 1219 | error = PTR_ERR(new_fl); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1220 | goto out; |
| 1221 | } |
| 1222 | |
| 1223 | break_time = 0; |
| 1224 | if (lease_break_time > 0) { |
| 1225 | break_time = jiffies + lease_break_time * HZ; |
| 1226 | if (break_time == 0) |
| 1227 | break_time++; /* so that 0 means no break time */ |
| 1228 | } |
| 1229 | |
| 1230 | for (fl = flock; fl && IS_LEASE(fl); fl = fl->fl_next) { |
| 1231 | if (fl->fl_type != future) { |
| 1232 | fl->fl_type = future; |
| 1233 | fl->fl_break_time = break_time; |
| 1234 | /* lease must have lmops break callback */ |
| 1235 | fl->fl_lmops->fl_break(fl); |
| 1236 | } |
| 1237 | } |
| 1238 | |
| 1239 | if (i_have_this_lease || (mode & O_NONBLOCK)) { |
| 1240 | error = -EWOULDBLOCK; |
| 1241 | goto out; |
| 1242 | } |
| 1243 | |
| 1244 | restart: |
| 1245 | break_time = flock->fl_break_time; |
| 1246 | if (break_time != 0) { |
| 1247 | break_time -= jiffies; |
| 1248 | if (break_time == 0) |
| 1249 | break_time++; |
| 1250 | } |
Matthew Wilcox | 4321e01 | 2008-01-14 21:28:30 -0700 | [diff] [blame] | 1251 | locks_insert_block(flock, new_fl); |
| 1252 | error = wait_event_interruptible_timeout(new_fl->fl_wait, |
| 1253 | !new_fl->fl_next, break_time); |
| 1254 | __locks_delete_block(new_fl); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1255 | if (error >= 0) { |
| 1256 | if (error == 0) |
| 1257 | time_out_leases(inode); |
| 1258 | /* Wait for the next lease that has not been broken yet */ |
| 1259 | for (flock = inode->i_flock; flock && IS_LEASE(flock); |
| 1260 | flock = flock->fl_next) { |
| 1261 | if (flock->fl_type & F_INPROGRESS) |
| 1262 | goto restart; |
| 1263 | } |
| 1264 | error = 0; |
| 1265 | } |
| 1266 | |
| 1267 | out: |
| 1268 | unlock_kernel(); |
J. Bruce Fields | e32b8ee | 2007-03-01 14:34:35 -0500 | [diff] [blame] | 1269 | if (!IS_ERR(new_fl)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1270 | locks_free_lock(new_fl); |
| 1271 | return error; |
| 1272 | } |
| 1273 | |
| 1274 | EXPORT_SYMBOL(__break_lease); |
| 1275 | |
| 1276 | /** |
Randy Dunlap | a6b9191 | 2008-03-19 17:01:00 -0700 | [diff] [blame] | 1277 | * lease_get_mtime - get the last modified time of an inode |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1278 | * @inode: the inode |
| 1279 | * @time: pointer to a timespec which will contain the last modified time |
| 1280 | * |
| 1281 | * This is to force NFS clients to flush their caches for files with |
| 1282 | * exclusive leases. The justification is that if someone has an |
Randy Dunlap | a6b9191 | 2008-03-19 17:01:00 -0700 | [diff] [blame] | 1283 | * exclusive lease, then they could be modifying it. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1284 | */ |
| 1285 | void lease_get_mtime(struct inode *inode, struct timespec *time) |
| 1286 | { |
| 1287 | struct file_lock *flock = inode->i_flock; |
| 1288 | if (flock && IS_LEASE(flock) && (flock->fl_type & F_WRLCK)) |
| 1289 | *time = current_fs_time(inode->i_sb); |
| 1290 | else |
| 1291 | *time = inode->i_mtime; |
| 1292 | } |
| 1293 | |
| 1294 | EXPORT_SYMBOL(lease_get_mtime); |
| 1295 | |
| 1296 | /** |
| 1297 | * fcntl_getlease - Enquire what lease is currently active |
| 1298 | * @filp: the file |
| 1299 | * |
| 1300 | * The value returned by this function will be one of |
| 1301 | * (if no lease break is pending): |
| 1302 | * |
| 1303 | * %F_RDLCK to indicate a shared lease is held. |
| 1304 | * |
| 1305 | * %F_WRLCK to indicate an exclusive lease is held. |
| 1306 | * |
| 1307 | * %F_UNLCK to indicate no lease is held. |
| 1308 | * |
| 1309 | * (if a lease break is pending): |
| 1310 | * |
| 1311 | * %F_RDLCK to indicate an exclusive lease needs to be |
| 1312 | * changed to a shared lease (or removed). |
| 1313 | * |
| 1314 | * %F_UNLCK to indicate the lease needs to be removed. |
| 1315 | * |
| 1316 | * XXX: sfr & willy disagree over whether F_INPROGRESS |
| 1317 | * should be returned to userspace. |
| 1318 | */ |
| 1319 | int fcntl_getlease(struct file *filp) |
| 1320 | { |
| 1321 | struct file_lock *fl; |
| 1322 | int type = F_UNLCK; |
| 1323 | |
| 1324 | lock_kernel(); |
Josef "Jeff" Sipek | 0f7fc9e | 2006-12-08 02:36:35 -0800 | [diff] [blame] | 1325 | time_out_leases(filp->f_path.dentry->d_inode); |
| 1326 | for (fl = filp->f_path.dentry->d_inode->i_flock; fl && IS_LEASE(fl); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1327 | fl = fl->fl_next) { |
| 1328 | if (fl->fl_file == filp) { |
| 1329 | type = fl->fl_type & ~F_INPROGRESS; |
| 1330 | break; |
| 1331 | } |
| 1332 | } |
| 1333 | unlock_kernel(); |
| 1334 | return type; |
| 1335 | } |
| 1336 | |
| 1337 | /** |
Christoph Hellwig | 0af1a45 | 2007-07-31 00:39:22 -0700 | [diff] [blame] | 1338 | * generic_setlease - sets a lease on an open file |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1339 | * @filp: file pointer |
| 1340 | * @arg: type of lease to obtain |
| 1341 | * @flp: input - file_lock to use, output - file_lock inserted |
| 1342 | * |
| 1343 | * The (input) flp->fl_lmops->fl_break function is required |
| 1344 | * by break_lease(). |
| 1345 | * |
| 1346 | * Called with kernel lock held. |
| 1347 | */ |
Christoph Hellwig | 0af1a45 | 2007-07-31 00:39:22 -0700 | [diff] [blame] | 1348 | int generic_setlease(struct file *filp, long arg, struct file_lock **flp) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1349 | { |
KAMBAROV, ZAUR | 7eaae28 | 2005-07-07 17:57:06 -0700 | [diff] [blame] | 1350 | struct file_lock *fl, **before, **my_before = NULL, *lease; |
Pavel Emelyanov | 85c5958 | 2007-09-20 12:45:02 +0400 | [diff] [blame] | 1351 | struct file_lock *new_fl = NULL; |
Josef "Jeff" Sipek | 0f7fc9e | 2006-12-08 02:36:35 -0800 | [diff] [blame] | 1352 | struct dentry *dentry = filp->f_path.dentry; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1353 | struct inode *inode = dentry->d_inode; |
| 1354 | int error, rdlease_count = 0, wrlease_count = 0; |
| 1355 | |
J. Bruce Fields | 6d5e8b0 | 2007-05-31 17:03:46 -0400 | [diff] [blame] | 1356 | if ((current->fsuid != inode->i_uid) && !capable(CAP_LEASE)) |
| 1357 | return -EACCES; |
| 1358 | if (!S_ISREG(inode->i_mode)) |
| 1359 | return -EINVAL; |
| 1360 | error = security_file_lock(filp, arg); |
| 1361 | if (error) |
| 1362 | return error; |
| 1363 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1364 | time_out_leases(inode); |
| 1365 | |
J. Bruce Fields | d2ab0b0c | 2007-06-30 12:40:32 -0400 | [diff] [blame] | 1366 | BUG_ON(!(*flp)->fl_lmops->fl_break); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1367 | |
KAMBAROV, ZAUR | 7eaae28 | 2005-07-07 17:57:06 -0700 | [diff] [blame] | 1368 | lease = *flp; |
| 1369 | |
David M. Richter | 288b2fd | 2008-04-23 16:29:00 -0400 | [diff] [blame] | 1370 | if (arg != F_UNLCK) { |
David M. Richter | 1908555 | 2008-04-23 16:29:01 -0400 | [diff] [blame] | 1371 | error = -ENOMEM; |
| 1372 | new_fl = locks_alloc_lock(); |
| 1373 | if (new_fl == NULL) |
| 1374 | goto out; |
| 1375 | |
David M. Richter | 288b2fd | 2008-04-23 16:29:00 -0400 | [diff] [blame] | 1376 | error = -EAGAIN; |
| 1377 | if ((arg == F_RDLCK) && (atomic_read(&inode->i_writecount) > 0)) |
| 1378 | goto out; |
| 1379 | if ((arg == F_WRLCK) |
| 1380 | && ((atomic_read(&dentry->d_count) > 1) |
| 1381 | || (atomic_read(&inode->i_count) > 1))) |
| 1382 | goto out; |
David M. Richter | 288b2fd | 2008-04-23 16:29:00 -0400 | [diff] [blame] | 1383 | } |
Pavel Emelyanov | 85c5958 | 2007-09-20 12:45:02 +0400 | [diff] [blame] | 1384 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1385 | /* |
| 1386 | * At this point, we know that if there is an exclusive |
| 1387 | * lease on this file, then we hold it on this filp |
| 1388 | * (otherwise our open of this file would have blocked). |
| 1389 | * And if we are trying to acquire an exclusive lease, |
| 1390 | * then the file is not open by anyone (including us) |
| 1391 | * except for this filp. |
| 1392 | */ |
| 1393 | for (before = &inode->i_flock; |
| 1394 | ((fl = *before) != NULL) && IS_LEASE(fl); |
| 1395 | before = &fl->fl_next) { |
| 1396 | if (lease->fl_lmops->fl_mylease(fl, lease)) |
| 1397 | my_before = before; |
| 1398 | else if (fl->fl_type == (F_INPROGRESS | F_UNLCK)) |
| 1399 | /* |
| 1400 | * Someone is in the process of opening this |
| 1401 | * file for writing so we may not take an |
| 1402 | * exclusive lease on it. |
| 1403 | */ |
| 1404 | wrlease_count++; |
| 1405 | else |
| 1406 | rdlease_count++; |
| 1407 | } |
| 1408 | |
David M. Richter | 5fcc60c | 2008-04-23 16:28:59 -0400 | [diff] [blame] | 1409 | error = -EAGAIN; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1410 | if ((arg == F_RDLCK && (wrlease_count > 0)) || |
| 1411 | (arg == F_WRLCK && ((rdlease_count + wrlease_count) > 0))) |
| 1412 | goto out; |
| 1413 | |
| 1414 | if (my_before != NULL) { |
Trond Myklebust | 75dff55 | 2006-05-07 23:02:42 -0400 | [diff] [blame] | 1415 | *flp = *my_before; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1416 | error = lease->fl_lmops->fl_change(my_before, arg); |
| 1417 | goto out; |
| 1418 | } |
| 1419 | |
| 1420 | error = 0; |
| 1421 | if (arg == F_UNLCK) |
| 1422 | goto out; |
| 1423 | |
| 1424 | error = -EINVAL; |
| 1425 | if (!leases_enable) |
| 1426 | goto out; |
| 1427 | |
Pavel Emelyanov | 85c5958 | 2007-09-20 12:45:02 +0400 | [diff] [blame] | 1428 | locks_copy_lock(new_fl, lease); |
| 1429 | locks_insert_lock(before, new_fl); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1430 | |
Pavel Emelyanov | 85c5958 | 2007-09-20 12:45:02 +0400 | [diff] [blame] | 1431 | *flp = new_fl; |
| 1432 | return 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1433 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1434 | out: |
Pavel Emelyanov | 85c5958 | 2007-09-20 12:45:02 +0400 | [diff] [blame] | 1435 | if (new_fl != NULL) |
| 1436 | locks_free_lock(new_fl); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1437 | return error; |
| 1438 | } |
Christoph Hellwig | 0af1a45 | 2007-07-31 00:39:22 -0700 | [diff] [blame] | 1439 | EXPORT_SYMBOL(generic_setlease); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1440 | |
| 1441 | /** |
J. Bruce Fields | a9933ce | 2007-06-07 17:09:49 -0400 | [diff] [blame] | 1442 | * vfs_setlease - sets a lease on an open file |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1443 | * @filp: file pointer |
| 1444 | * @arg: type of lease to obtain |
| 1445 | * @lease: file_lock to use |
| 1446 | * |
| 1447 | * Call this to establish a lease on the file. |
J. Bruce Fields | f9ffed2 | 2006-11-14 15:51:40 -0500 | [diff] [blame] | 1448 | * The (*lease)->fl_lmops->fl_break operation must be set; if not, |
| 1449 | * break_lease will oops! |
| 1450 | * |
| 1451 | * This will call the filesystem's setlease file method, if |
| 1452 | * defined. Note that there is no getlease method; instead, the |
| 1453 | * filesystem setlease method should call back to setlease() to |
| 1454 | * add a lease to the inode's lease list, where fcntl_getlease() can |
| 1455 | * find it. Since fcntl_getlease() only reports whether the current |
| 1456 | * task holds a lease, a cluster filesystem need only do this for |
| 1457 | * leases held by processes on this node. |
| 1458 | * |
| 1459 | * There is also no break_lease method; filesystems that |
| 1460 | * handle their own leases shoud break leases themselves from the |
| 1461 | * filesystem's open, create, and (on truncate) setattr methods. |
| 1462 | * |
| 1463 | * Warning: the only current setlease methods exist only to disable |
| 1464 | * leases in certain cases. More vfs changes may be required to |
| 1465 | * allow a full filesystem lease implementation. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1466 | */ |
| 1467 | |
J. Bruce Fields | a9933ce | 2007-06-07 17:09:49 -0400 | [diff] [blame] | 1468 | int vfs_setlease(struct file *filp, long arg, struct file_lock **lease) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1469 | { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1470 | int error; |
| 1471 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1472 | lock_kernel(); |
J. Bruce Fields | f9ffed2 | 2006-11-14 15:51:40 -0500 | [diff] [blame] | 1473 | if (filp->f_op && filp->f_op->setlease) |
| 1474 | error = filp->f_op->setlease(filp, arg, lease); |
| 1475 | else |
Christoph Hellwig | 0af1a45 | 2007-07-31 00:39:22 -0700 | [diff] [blame] | 1476 | error = generic_setlease(filp, arg, lease); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1477 | unlock_kernel(); |
| 1478 | |
| 1479 | return error; |
| 1480 | } |
J. Bruce Fields | a9933ce | 2007-06-07 17:09:49 -0400 | [diff] [blame] | 1481 | EXPORT_SYMBOL_GPL(vfs_setlease); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1482 | |
| 1483 | /** |
| 1484 | * fcntl_setlease - sets a lease on an open file |
| 1485 | * @fd: open file descriptor |
| 1486 | * @filp: file pointer |
| 1487 | * @arg: type of lease to obtain |
| 1488 | * |
| 1489 | * Call this fcntl to establish a lease on the file. |
| 1490 | * Note that you also need to call %F_SETSIG to |
| 1491 | * receive a signal when the lease is broken. |
| 1492 | */ |
| 1493 | int fcntl_setlease(unsigned int fd, struct file *filp, long arg) |
| 1494 | { |
| 1495 | struct file_lock fl, *flp = &fl; |
David M. Richter | 9d91cdc | 2008-04-23 16:29:02 -0400 | [diff] [blame] | 1496 | struct inode *inode = filp->f_path.dentry->d_inode; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1497 | int error; |
| 1498 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1499 | locks_init_lock(&fl); |
| 1500 | error = lease_init(filp, arg, &fl); |
| 1501 | if (error) |
| 1502 | return error; |
| 1503 | |
| 1504 | lock_kernel(); |
| 1505 | |
J. Bruce Fields | a9933ce | 2007-06-07 17:09:49 -0400 | [diff] [blame] | 1506 | error = vfs_setlease(filp, arg, &flp); |
J. Bruce Fields | dc15ae1 | 2005-11-10 19:08:00 -0500 | [diff] [blame] | 1507 | if (error || arg == F_UNLCK) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1508 | goto out_unlock; |
| 1509 | |
| 1510 | error = fasync_helper(fd, filp, 1, &flp->fl_fasync); |
| 1511 | if (error < 0) { |
J. Bruce Fields | 6d5e8b0 | 2007-05-31 17:03:46 -0400 | [diff] [blame] | 1512 | /* remove lease just inserted by setlease */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1513 | flp->fl_type = F_UNLCK | F_INPROGRESS; |
J. Bruce Fields | 6d5e8b0 | 2007-05-31 17:03:46 -0400 | [diff] [blame] | 1514 | flp->fl_break_time = jiffies - 10; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1515 | time_out_leases(inode); |
| 1516 | goto out_unlock; |
| 1517 | } |
| 1518 | |
Eric W. Biederman | 609d7fa | 2006-10-02 02:17:15 -0700 | [diff] [blame] | 1519 | error = __f_setown(filp, task_pid(current), PIDTYPE_PID, 0); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1520 | out_unlock: |
| 1521 | unlock_kernel(); |
| 1522 | return error; |
| 1523 | } |
| 1524 | |
| 1525 | /** |
| 1526 | * flock_lock_file_wait - Apply a FLOCK-style lock to a file |
| 1527 | * @filp: The file to apply the lock to |
| 1528 | * @fl: The lock to be applied |
| 1529 | * |
| 1530 | * Add a FLOCK style lock to a file. |
| 1531 | */ |
| 1532 | int flock_lock_file_wait(struct file *filp, struct file_lock *fl) |
| 1533 | { |
| 1534 | int error; |
| 1535 | might_sleep(); |
| 1536 | for (;;) { |
| 1537 | error = flock_lock_file(filp, fl); |
| 1538 | if ((error != -EAGAIN) || !(fl->fl_flags & FL_SLEEP)) |
| 1539 | break; |
| 1540 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_next); |
| 1541 | if (!error) |
| 1542 | continue; |
| 1543 | |
| 1544 | locks_delete_block(fl); |
| 1545 | break; |
| 1546 | } |
| 1547 | return error; |
| 1548 | } |
| 1549 | |
| 1550 | EXPORT_SYMBOL(flock_lock_file_wait); |
| 1551 | |
| 1552 | /** |
| 1553 | * sys_flock: - flock() system call. |
| 1554 | * @fd: the file descriptor to lock. |
| 1555 | * @cmd: the type of lock to apply. |
| 1556 | * |
| 1557 | * Apply a %FL_FLOCK style lock to an open file descriptor. |
| 1558 | * The @cmd can be one of |
| 1559 | * |
| 1560 | * %LOCK_SH -- a shared lock. |
| 1561 | * |
| 1562 | * %LOCK_EX -- an exclusive lock. |
| 1563 | * |
| 1564 | * %LOCK_UN -- remove an existing lock. |
| 1565 | * |
| 1566 | * %LOCK_MAND -- a `mandatory' flock. This exists to emulate Windows Share Modes. |
| 1567 | * |
| 1568 | * %LOCK_MAND can be combined with %LOCK_READ or %LOCK_WRITE to allow other |
| 1569 | * processes read and write access respectively. |
| 1570 | */ |
| 1571 | asmlinkage long sys_flock(unsigned int fd, unsigned int cmd) |
| 1572 | { |
| 1573 | struct file *filp; |
| 1574 | struct file_lock *lock; |
| 1575 | int can_sleep, unlock; |
| 1576 | int error; |
| 1577 | |
| 1578 | error = -EBADF; |
| 1579 | filp = fget(fd); |
| 1580 | if (!filp) |
| 1581 | goto out; |
| 1582 | |
| 1583 | can_sleep = !(cmd & LOCK_NB); |
| 1584 | cmd &= ~LOCK_NB; |
| 1585 | unlock = (cmd == LOCK_UN); |
| 1586 | |
| 1587 | if (!unlock && !(cmd & LOCK_MAND) && !(filp->f_mode & 3)) |
| 1588 | goto out_putf; |
| 1589 | |
| 1590 | error = flock_make_lock(filp, &lock, cmd); |
| 1591 | if (error) |
| 1592 | goto out_putf; |
| 1593 | if (can_sleep) |
| 1594 | lock->fl_flags |= FL_SLEEP; |
| 1595 | |
| 1596 | error = security_file_lock(filp, cmd); |
| 1597 | if (error) |
| 1598 | goto out_free; |
| 1599 | |
| 1600 | if (filp->f_op && filp->f_op->flock) |
| 1601 | error = filp->f_op->flock(filp, |
| 1602 | (can_sleep) ? F_SETLKW : F_SETLK, |
| 1603 | lock); |
| 1604 | else |
| 1605 | error = flock_lock_file_wait(filp, lock); |
| 1606 | |
| 1607 | out_free: |
Trond Myklebust | 993dfa8 | 2006-03-31 02:30:55 -0800 | [diff] [blame] | 1608 | locks_free_lock(lock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1609 | |
| 1610 | out_putf: |
| 1611 | fput(filp); |
| 1612 | out: |
| 1613 | return error; |
| 1614 | } |
| 1615 | |
J. Bruce Fields | 3ee17ab | 2007-02-21 00:58:50 -0500 | [diff] [blame] | 1616 | /** |
| 1617 | * vfs_test_lock - test file byte range lock |
| 1618 | * @filp: The file to test lock for |
J. Bruce Fields | 6924c55 | 2007-05-11 16:22:50 -0400 | [diff] [blame] | 1619 | * @fl: The lock to test; also used to hold result |
J. Bruce Fields | 3ee17ab | 2007-02-21 00:58:50 -0500 | [diff] [blame] | 1620 | * |
| 1621 | * Returns -ERRNO on failure. Indicates presence of conflicting lock by |
| 1622 | * setting conf->fl_type to something other than F_UNLCK. |
| 1623 | */ |
| 1624 | int vfs_test_lock(struct file *filp, struct file_lock *fl) |
| 1625 | { |
| 1626 | if (filp->f_op && filp->f_op->lock) |
| 1627 | return filp->f_op->lock(filp, F_GETLK, fl); |
| 1628 | posix_test_lock(filp, fl); |
| 1629 | return 0; |
| 1630 | } |
| 1631 | EXPORT_SYMBOL_GPL(vfs_test_lock); |
| 1632 | |
J. Bruce Fields | c2fa1b8 | 2007-02-20 16:10:11 -0500 | [diff] [blame] | 1633 | static int posix_lock_to_flock(struct flock *flock, struct file_lock *fl) |
| 1634 | { |
| 1635 | flock->l_pid = fl->fl_pid; |
| 1636 | #if BITS_PER_LONG == 32 |
| 1637 | /* |
| 1638 | * Make sure we can represent the posix lock via |
| 1639 | * legacy 32bit flock. |
| 1640 | */ |
| 1641 | if (fl->fl_start > OFFT_OFFSET_MAX) |
| 1642 | return -EOVERFLOW; |
| 1643 | if (fl->fl_end != OFFSET_MAX && fl->fl_end > OFFT_OFFSET_MAX) |
| 1644 | return -EOVERFLOW; |
| 1645 | #endif |
| 1646 | flock->l_start = fl->fl_start; |
| 1647 | flock->l_len = fl->fl_end == OFFSET_MAX ? 0 : |
| 1648 | fl->fl_end - fl->fl_start + 1; |
| 1649 | flock->l_whence = 0; |
J. Bruce Fields | 129a84d | 2007-05-10 18:38:43 -0400 | [diff] [blame] | 1650 | flock->l_type = fl->fl_type; |
J. Bruce Fields | c2fa1b8 | 2007-02-20 16:10:11 -0500 | [diff] [blame] | 1651 | return 0; |
| 1652 | } |
| 1653 | |
| 1654 | #if BITS_PER_LONG == 32 |
| 1655 | static void posix_lock_to_flock64(struct flock64 *flock, struct file_lock *fl) |
| 1656 | { |
| 1657 | flock->l_pid = fl->fl_pid; |
| 1658 | flock->l_start = fl->fl_start; |
| 1659 | flock->l_len = fl->fl_end == OFFSET_MAX ? 0 : |
| 1660 | fl->fl_end - fl->fl_start + 1; |
| 1661 | flock->l_whence = 0; |
| 1662 | flock->l_type = fl->fl_type; |
| 1663 | } |
| 1664 | #endif |
| 1665 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1666 | /* Report the first existing lock that would conflict with l. |
| 1667 | * This implements the F_GETLK command of fcntl(). |
| 1668 | */ |
| 1669 | int fcntl_getlk(struct file *filp, struct flock __user *l) |
| 1670 | { |
Marc Eshel | 9d6a8c5 | 2007-02-21 00:55:18 -0500 | [diff] [blame] | 1671 | struct file_lock file_lock; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1672 | struct flock flock; |
| 1673 | int error; |
| 1674 | |
| 1675 | error = -EFAULT; |
| 1676 | if (copy_from_user(&flock, l, sizeof(flock))) |
| 1677 | goto out; |
| 1678 | error = -EINVAL; |
| 1679 | if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK)) |
| 1680 | goto out; |
| 1681 | |
| 1682 | error = flock_to_posix_lock(filp, &file_lock, &flock); |
| 1683 | if (error) |
| 1684 | goto out; |
| 1685 | |
J. Bruce Fields | 3ee17ab | 2007-02-21 00:58:50 -0500 | [diff] [blame] | 1686 | error = vfs_test_lock(filp, &file_lock); |
| 1687 | if (error) |
| 1688 | goto out; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1689 | |
Marc Eshel | 9d6a8c5 | 2007-02-21 00:55:18 -0500 | [diff] [blame] | 1690 | flock.l_type = file_lock.fl_type; |
| 1691 | if (file_lock.fl_type != F_UNLCK) { |
| 1692 | error = posix_lock_to_flock(&flock, &file_lock); |
J. Bruce Fields | c2fa1b8 | 2007-02-20 16:10:11 -0500 | [diff] [blame] | 1693 | if (error) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1694 | goto out; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1695 | } |
| 1696 | error = -EFAULT; |
| 1697 | if (!copy_to_user(l, &flock, sizeof(flock))) |
| 1698 | error = 0; |
| 1699 | out: |
| 1700 | return error; |
| 1701 | } |
| 1702 | |
Marc Eshel | 7723ec9 | 2007-01-18 15:08:55 -0500 | [diff] [blame] | 1703 | /** |
| 1704 | * vfs_lock_file - file byte range lock |
| 1705 | * @filp: The file to apply the lock to |
| 1706 | * @cmd: type of locking operation (F_SETLK, F_GETLK, etc.) |
| 1707 | * @fl: The lock to be applied |
Marc Eshel | 150b393 | 2007-01-18 16:15:35 -0500 | [diff] [blame] | 1708 | * @conf: Place to return a copy of the conflicting lock, if found. |
| 1709 | * |
| 1710 | * A caller that doesn't care about the conflicting lock may pass NULL |
| 1711 | * as the final argument. |
| 1712 | * |
| 1713 | * If the filesystem defines a private ->lock() method, then @conf will |
| 1714 | * be left unchanged; so a caller that cares should initialize it to |
| 1715 | * some acceptable default. |
Marc Eshel | 2beb661 | 2006-12-05 23:31:28 -0500 | [diff] [blame] | 1716 | * |
| 1717 | * To avoid blocking kernel daemons, such as lockd, that need to acquire POSIX |
| 1718 | * locks, the ->lock() interface may return asynchronously, before the lock has |
| 1719 | * been granted or denied by the underlying filesystem, if (and only if) |
| 1720 | * fl_grant is set. Callers expecting ->lock() to return asynchronously |
| 1721 | * will only use F_SETLK, not F_SETLKW; they will set FL_SLEEP if (and only if) |
| 1722 | * the request is for a blocking lock. When ->lock() does return asynchronously, |
| 1723 | * it must return -EINPROGRESS, and call ->fl_grant() when the lock |
| 1724 | * request completes. |
| 1725 | * If the request is for non-blocking lock the file system should return |
| 1726 | * -EINPROGRESS then try to get the lock and call the callback routine with |
| 1727 | * the result. If the request timed out the callback routine will return a |
| 1728 | * nonzero return code and the file system should release the lock. The file |
| 1729 | * system is also responsible to keep a corresponding posix lock when it |
| 1730 | * grants a lock so the VFS can find out which locks are locally held and do |
| 1731 | * the correct lock cleanup when required. |
| 1732 | * The underlying filesystem must not drop the kernel lock or call |
| 1733 | * ->fl_grant() before returning to the caller with a -EINPROGRESS |
| 1734 | * return code. |
Marc Eshel | 7723ec9 | 2007-01-18 15:08:55 -0500 | [diff] [blame] | 1735 | */ |
Marc Eshel | 150b393 | 2007-01-18 16:15:35 -0500 | [diff] [blame] | 1736 | int vfs_lock_file(struct file *filp, unsigned int cmd, struct file_lock *fl, struct file_lock *conf) |
Marc Eshel | 7723ec9 | 2007-01-18 15:08:55 -0500 | [diff] [blame] | 1737 | { |
| 1738 | if (filp->f_op && filp->f_op->lock) |
| 1739 | return filp->f_op->lock(filp, cmd, fl); |
| 1740 | else |
Marc Eshel | 150b393 | 2007-01-18 16:15:35 -0500 | [diff] [blame] | 1741 | return posix_lock_file(filp, fl, conf); |
Marc Eshel | 7723ec9 | 2007-01-18 15:08:55 -0500 | [diff] [blame] | 1742 | } |
| 1743 | EXPORT_SYMBOL_GPL(vfs_lock_file); |
| 1744 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1745 | /* Apply the lock described by l to an open file descriptor. |
| 1746 | * This implements both the F_SETLK and F_SETLKW commands of fcntl(). |
| 1747 | */ |
Peter Staubach | c293621 | 2005-07-27 11:45:09 -0700 | [diff] [blame] | 1748 | int fcntl_setlk(unsigned int fd, struct file *filp, unsigned int cmd, |
| 1749 | struct flock __user *l) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1750 | { |
| 1751 | struct file_lock *file_lock = locks_alloc_lock(); |
| 1752 | struct flock flock; |
| 1753 | struct inode *inode; |
| 1754 | int error; |
| 1755 | |
| 1756 | if (file_lock == NULL) |
| 1757 | return -ENOLCK; |
| 1758 | |
| 1759 | /* |
| 1760 | * This might block, so we do it before checking the inode. |
| 1761 | */ |
| 1762 | error = -EFAULT; |
| 1763 | if (copy_from_user(&flock, l, sizeof(flock))) |
| 1764 | goto out; |
| 1765 | |
Josef "Jeff" Sipek | 0f7fc9e | 2006-12-08 02:36:35 -0800 | [diff] [blame] | 1766 | inode = filp->f_path.dentry->d_inode; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1767 | |
| 1768 | /* Don't allow mandatory locks on files that may be memory mapped |
| 1769 | * and shared. |
| 1770 | */ |
Pavel Emelyanov | a16877c | 2007-10-01 14:41:11 -0700 | [diff] [blame] | 1771 | if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1772 | error = -EAGAIN; |
| 1773 | goto out; |
| 1774 | } |
| 1775 | |
Peter Staubach | c293621 | 2005-07-27 11:45:09 -0700 | [diff] [blame] | 1776 | again: |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1777 | error = flock_to_posix_lock(filp, file_lock, &flock); |
| 1778 | if (error) |
| 1779 | goto out; |
| 1780 | if (cmd == F_SETLKW) { |
| 1781 | file_lock->fl_flags |= FL_SLEEP; |
| 1782 | } |
| 1783 | |
| 1784 | error = -EBADF; |
| 1785 | switch (flock.l_type) { |
| 1786 | case F_RDLCK: |
| 1787 | if (!(filp->f_mode & FMODE_READ)) |
| 1788 | goto out; |
| 1789 | break; |
| 1790 | case F_WRLCK: |
| 1791 | if (!(filp->f_mode & FMODE_WRITE)) |
| 1792 | goto out; |
| 1793 | break; |
| 1794 | case F_UNLCK: |
| 1795 | break; |
| 1796 | default: |
| 1797 | error = -EINVAL; |
| 1798 | goto out; |
| 1799 | } |
| 1800 | |
| 1801 | error = security_file_lock(filp, file_lock->fl_type); |
| 1802 | if (error) |
| 1803 | goto out; |
| 1804 | |
J. Bruce Fields | 19e729a | 2008-04-14 15:03:02 -0400 | [diff] [blame] | 1805 | if (filp->f_op && filp->f_op->lock != NULL) |
| 1806 | error = filp->f_op->lock(filp, cmd, file_lock); |
| 1807 | else { |
| 1808 | for (;;) { |
| 1809 | error = posix_lock_file(filp, file_lock, NULL); |
| 1810 | if (error != -EAGAIN || cmd == F_SETLK) |
| 1811 | break; |
| 1812 | error = wait_event_interruptible(file_lock->fl_wait, |
| 1813 | !file_lock->fl_next); |
| 1814 | if (!error) |
| 1815 | continue; |
Marc Eshel | 7723ec9 | 2007-01-18 15:08:55 -0500 | [diff] [blame] | 1816 | |
J. Bruce Fields | 19e729a | 2008-04-14 15:03:02 -0400 | [diff] [blame] | 1817 | locks_delete_block(file_lock); |
| 1818 | break; |
| 1819 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1820 | } |
| 1821 | |
Peter Staubach | c293621 | 2005-07-27 11:45:09 -0700 | [diff] [blame] | 1822 | /* |
| 1823 | * Attempt to detect a close/fcntl race and recover by |
| 1824 | * releasing the lock that was just acquired. |
| 1825 | */ |
| 1826 | if (!error && fcheck(fd) != filp && flock.l_type != F_UNLCK) { |
| 1827 | flock.l_type = F_UNLCK; |
| 1828 | goto again; |
| 1829 | } |
| 1830 | |
| 1831 | out: |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1832 | locks_free_lock(file_lock); |
| 1833 | return error; |
| 1834 | } |
| 1835 | |
| 1836 | #if BITS_PER_LONG == 32 |
| 1837 | /* Report the first existing lock that would conflict with l. |
| 1838 | * This implements the F_GETLK command of fcntl(). |
| 1839 | */ |
| 1840 | int fcntl_getlk64(struct file *filp, struct flock64 __user *l) |
| 1841 | { |
Marc Eshel | 9d6a8c5 | 2007-02-21 00:55:18 -0500 | [diff] [blame] | 1842 | struct file_lock file_lock; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1843 | struct flock64 flock; |
| 1844 | int error; |
| 1845 | |
| 1846 | error = -EFAULT; |
| 1847 | if (copy_from_user(&flock, l, sizeof(flock))) |
| 1848 | goto out; |
| 1849 | error = -EINVAL; |
| 1850 | if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK)) |
| 1851 | goto out; |
| 1852 | |
| 1853 | error = flock64_to_posix_lock(filp, &file_lock, &flock); |
| 1854 | if (error) |
| 1855 | goto out; |
| 1856 | |
J. Bruce Fields | 3ee17ab | 2007-02-21 00:58:50 -0500 | [diff] [blame] | 1857 | error = vfs_test_lock(filp, &file_lock); |
| 1858 | if (error) |
| 1859 | goto out; |
| 1860 | |
Marc Eshel | 9d6a8c5 | 2007-02-21 00:55:18 -0500 | [diff] [blame] | 1861 | flock.l_type = file_lock.fl_type; |
| 1862 | if (file_lock.fl_type != F_UNLCK) |
| 1863 | posix_lock_to_flock64(&flock, &file_lock); |
| 1864 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1865 | error = -EFAULT; |
| 1866 | if (!copy_to_user(l, &flock, sizeof(flock))) |
| 1867 | error = 0; |
| 1868 | |
| 1869 | out: |
| 1870 | return error; |
| 1871 | } |
| 1872 | |
| 1873 | /* Apply the lock described by l to an open file descriptor. |
| 1874 | * This implements both the F_SETLK and F_SETLKW commands of fcntl(). |
| 1875 | */ |
Peter Staubach | c293621 | 2005-07-27 11:45:09 -0700 | [diff] [blame] | 1876 | int fcntl_setlk64(unsigned int fd, struct file *filp, unsigned int cmd, |
| 1877 | struct flock64 __user *l) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1878 | { |
| 1879 | struct file_lock *file_lock = locks_alloc_lock(); |
| 1880 | struct flock64 flock; |
| 1881 | struct inode *inode; |
| 1882 | int error; |
| 1883 | |
| 1884 | if (file_lock == NULL) |
| 1885 | return -ENOLCK; |
| 1886 | |
| 1887 | /* |
| 1888 | * This might block, so we do it before checking the inode. |
| 1889 | */ |
| 1890 | error = -EFAULT; |
| 1891 | if (copy_from_user(&flock, l, sizeof(flock))) |
| 1892 | goto out; |
| 1893 | |
Josef "Jeff" Sipek | 0f7fc9e | 2006-12-08 02:36:35 -0800 | [diff] [blame] | 1894 | inode = filp->f_path.dentry->d_inode; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1895 | |
| 1896 | /* Don't allow mandatory locks on files that may be memory mapped |
| 1897 | * and shared. |
| 1898 | */ |
Pavel Emelyanov | a16877c | 2007-10-01 14:41:11 -0700 | [diff] [blame] | 1899 | if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1900 | error = -EAGAIN; |
| 1901 | goto out; |
| 1902 | } |
| 1903 | |
Peter Staubach | c293621 | 2005-07-27 11:45:09 -0700 | [diff] [blame] | 1904 | again: |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1905 | error = flock64_to_posix_lock(filp, file_lock, &flock); |
| 1906 | if (error) |
| 1907 | goto out; |
| 1908 | if (cmd == F_SETLKW64) { |
| 1909 | file_lock->fl_flags |= FL_SLEEP; |
| 1910 | } |
| 1911 | |
| 1912 | error = -EBADF; |
| 1913 | switch (flock.l_type) { |
| 1914 | case F_RDLCK: |
| 1915 | if (!(filp->f_mode & FMODE_READ)) |
| 1916 | goto out; |
| 1917 | break; |
| 1918 | case F_WRLCK: |
| 1919 | if (!(filp->f_mode & FMODE_WRITE)) |
| 1920 | goto out; |
| 1921 | break; |
| 1922 | case F_UNLCK: |
| 1923 | break; |
| 1924 | default: |
| 1925 | error = -EINVAL; |
| 1926 | goto out; |
| 1927 | } |
| 1928 | |
| 1929 | error = security_file_lock(filp, file_lock->fl_type); |
| 1930 | if (error) |
| 1931 | goto out; |
| 1932 | |
J. Bruce Fields | 19e729a | 2008-04-14 15:03:02 -0400 | [diff] [blame] | 1933 | if (filp->f_op && filp->f_op->lock != NULL) |
| 1934 | error = filp->f_op->lock(filp, cmd, file_lock); |
| 1935 | else { |
| 1936 | for (;;) { |
| 1937 | error = posix_lock_file(filp, file_lock, NULL); |
| 1938 | if (error != -EAGAIN || cmd == F_SETLK64) |
| 1939 | break; |
| 1940 | error = wait_event_interruptible(file_lock->fl_wait, |
| 1941 | !file_lock->fl_next); |
| 1942 | if (!error) |
| 1943 | continue; |
Marc Eshel | 7723ec9 | 2007-01-18 15:08:55 -0500 | [diff] [blame] | 1944 | |
J. Bruce Fields | 19e729a | 2008-04-14 15:03:02 -0400 | [diff] [blame] | 1945 | locks_delete_block(file_lock); |
| 1946 | break; |
| 1947 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1948 | } |
| 1949 | |
Peter Staubach | c293621 | 2005-07-27 11:45:09 -0700 | [diff] [blame] | 1950 | /* |
| 1951 | * Attempt to detect a close/fcntl race and recover by |
| 1952 | * releasing the lock that was just acquired. |
| 1953 | */ |
| 1954 | if (!error && fcheck(fd) != filp && flock.l_type != F_UNLCK) { |
| 1955 | flock.l_type = F_UNLCK; |
| 1956 | goto again; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1957 | } |
| 1958 | |
| 1959 | out: |
| 1960 | locks_free_lock(file_lock); |
| 1961 | return error; |
| 1962 | } |
| 1963 | #endif /* BITS_PER_LONG == 32 */ |
| 1964 | |
| 1965 | /* |
| 1966 | * This function is called when the file is being removed |
| 1967 | * from the task's fd array. POSIX locks belonging to this task |
| 1968 | * are deleted at this time. |
| 1969 | */ |
| 1970 | void locks_remove_posix(struct file *filp, fl_owner_t owner) |
| 1971 | { |
Miklos Szeredi | ff7b86b | 2006-06-23 02:05:11 -0700 | [diff] [blame] | 1972 | struct file_lock lock; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1973 | |
| 1974 | /* |
| 1975 | * If there are no locks held on this file, we don't need to call |
| 1976 | * posix_lock_file(). Another process could be setting a lock on this |
| 1977 | * file at the same time, but we wouldn't remove that lock anyway. |
| 1978 | */ |
Josef "Jeff" Sipek | 0f7fc9e | 2006-12-08 02:36:35 -0800 | [diff] [blame] | 1979 | if (!filp->f_path.dentry->d_inode->i_flock) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1980 | return; |
| 1981 | |
| 1982 | lock.fl_type = F_UNLCK; |
Miklos Szeredi | 75e1fcc | 2006-06-23 02:05:12 -0700 | [diff] [blame] | 1983 | lock.fl_flags = FL_POSIX | FL_CLOSE; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1984 | lock.fl_start = 0; |
| 1985 | lock.fl_end = OFFSET_MAX; |
| 1986 | lock.fl_owner = owner; |
| 1987 | lock.fl_pid = current->tgid; |
| 1988 | lock.fl_file = filp; |
| 1989 | lock.fl_ops = NULL; |
| 1990 | lock.fl_lmops = NULL; |
| 1991 | |
Marc Eshel | 150b393 | 2007-01-18 16:15:35 -0500 | [diff] [blame] | 1992 | vfs_lock_file(filp, F_SETLK, &lock, NULL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1993 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1994 | if (lock.fl_ops && lock.fl_ops->fl_release_private) |
| 1995 | lock.fl_ops->fl_release_private(&lock); |
| 1996 | } |
| 1997 | |
| 1998 | EXPORT_SYMBOL(locks_remove_posix); |
| 1999 | |
| 2000 | /* |
| 2001 | * This function is called on the last close of an open file. |
| 2002 | */ |
| 2003 | void locks_remove_flock(struct file *filp) |
| 2004 | { |
Josef "Jeff" Sipek | 0f7fc9e | 2006-12-08 02:36:35 -0800 | [diff] [blame] | 2005 | struct inode * inode = filp->f_path.dentry->d_inode; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2006 | struct file_lock *fl; |
| 2007 | struct file_lock **before; |
| 2008 | |
| 2009 | if (!inode->i_flock) |
| 2010 | return; |
| 2011 | |
| 2012 | if (filp->f_op && filp->f_op->flock) { |
| 2013 | struct file_lock fl = { |
| 2014 | .fl_pid = current->tgid, |
| 2015 | .fl_file = filp, |
| 2016 | .fl_flags = FL_FLOCK, |
| 2017 | .fl_type = F_UNLCK, |
| 2018 | .fl_end = OFFSET_MAX, |
| 2019 | }; |
| 2020 | filp->f_op->flock(filp, F_SETLKW, &fl); |
Trond Myklebust | 80fec4c | 2005-06-22 17:16:31 +0000 | [diff] [blame] | 2021 | if (fl.fl_ops && fl.fl_ops->fl_release_private) |
| 2022 | fl.fl_ops->fl_release_private(&fl); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2023 | } |
| 2024 | |
| 2025 | lock_kernel(); |
| 2026 | before = &inode->i_flock; |
| 2027 | |
| 2028 | while ((fl = *before) != NULL) { |
| 2029 | if (fl->fl_file == filp) { |
Peter Staubach | c293621 | 2005-07-27 11:45:09 -0700 | [diff] [blame] | 2030 | if (IS_FLOCK(fl)) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2031 | locks_delete_lock(before); |
| 2032 | continue; |
| 2033 | } |
| 2034 | if (IS_LEASE(fl)) { |
| 2035 | lease_modify(before, F_UNLCK); |
| 2036 | continue; |
| 2037 | } |
| 2038 | /* What? */ |
| 2039 | BUG(); |
| 2040 | } |
| 2041 | before = &fl->fl_next; |
| 2042 | } |
| 2043 | unlock_kernel(); |
| 2044 | } |
| 2045 | |
| 2046 | /** |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2047 | * posix_unblock_lock - stop waiting for a file lock |
| 2048 | * @filp: how the file was opened |
| 2049 | * @waiter: the lock which was waiting |
| 2050 | * |
| 2051 | * lockd needs to block waiting for locks. |
| 2052 | */ |
J. Bruce Fields | 64a318e | 2006-01-03 09:55:46 +0100 | [diff] [blame] | 2053 | int |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2054 | posix_unblock_lock(struct file *filp, struct file_lock *waiter) |
| 2055 | { |
J. Bruce Fields | 64a318e | 2006-01-03 09:55:46 +0100 | [diff] [blame] | 2056 | int status = 0; |
| 2057 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2058 | lock_kernel(); |
J. Bruce Fields | 5996a29 | 2006-01-03 09:55:44 +0100 | [diff] [blame] | 2059 | if (waiter->fl_next) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2060 | __locks_delete_block(waiter); |
J. Bruce Fields | 64a318e | 2006-01-03 09:55:46 +0100 | [diff] [blame] | 2061 | else |
| 2062 | status = -ENOENT; |
J. Bruce Fields | 5996a29 | 2006-01-03 09:55:44 +0100 | [diff] [blame] | 2063 | unlock_kernel(); |
J. Bruce Fields | 64a318e | 2006-01-03 09:55:46 +0100 | [diff] [blame] | 2064 | return status; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2065 | } |
| 2066 | |
| 2067 | EXPORT_SYMBOL(posix_unblock_lock); |
| 2068 | |
Marc Eshel | 9b9d2ab | 2007-01-18 17:52:58 -0500 | [diff] [blame] | 2069 | /** |
| 2070 | * vfs_cancel_lock - file byte range unblock lock |
| 2071 | * @filp: The file to apply the unblock to |
| 2072 | * @fl: The lock to be unblocked |
| 2073 | * |
| 2074 | * Used by lock managers to cancel blocked requests |
| 2075 | */ |
| 2076 | int vfs_cancel_lock(struct file *filp, struct file_lock *fl) |
| 2077 | { |
| 2078 | if (filp->f_op && filp->f_op->lock) |
| 2079 | return filp->f_op->lock(filp, F_CANCELLK, fl); |
| 2080 | return 0; |
| 2081 | } |
| 2082 | |
| 2083 | EXPORT_SYMBOL_GPL(vfs_cancel_lock); |
| 2084 | |
Pavel Emelyanov | 7f8ada9 | 2007-10-01 14:41:15 -0700 | [diff] [blame] | 2085 | #ifdef CONFIG_PROC_FS |
| 2086 | #include <linux/seq_file.h> |
| 2087 | |
| 2088 | static void lock_get_status(struct seq_file *f, struct file_lock *fl, |
| 2089 | int id, char *pfx) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2090 | { |
| 2091 | struct inode *inode = NULL; |
Vitaliy Gusev | ab1f161 | 2008-01-17 00:07:08 +0000 | [diff] [blame] | 2092 | unsigned int fl_pid; |
| 2093 | |
| 2094 | if (fl->fl_nspid) |
Pavel Emelyanov | 6c5f3e7 | 2008-02-08 04:19:20 -0800 | [diff] [blame] | 2095 | fl_pid = pid_vnr(fl->fl_nspid); |
Vitaliy Gusev | ab1f161 | 2008-01-17 00:07:08 +0000 | [diff] [blame] | 2096 | else |
| 2097 | fl_pid = fl->fl_pid; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2098 | |
| 2099 | if (fl->fl_file != NULL) |
Josef "Jeff" Sipek | 0f7fc9e | 2006-12-08 02:36:35 -0800 | [diff] [blame] | 2100 | inode = fl->fl_file->f_path.dentry->d_inode; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2101 | |
Pavel Emelyanov | 7f8ada9 | 2007-10-01 14:41:15 -0700 | [diff] [blame] | 2102 | seq_printf(f, "%d:%s ", id, pfx); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2103 | if (IS_POSIX(fl)) { |
Pavel Emelyanov | 7f8ada9 | 2007-10-01 14:41:15 -0700 | [diff] [blame] | 2104 | seq_printf(f, "%6s %s ", |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2105 | (fl->fl_flags & FL_ACCESS) ? "ACCESS" : "POSIX ", |
| 2106 | (inode == NULL) ? "*NOINODE*" : |
Pavel Emelyanov | a16877c | 2007-10-01 14:41:11 -0700 | [diff] [blame] | 2107 | mandatory_lock(inode) ? "MANDATORY" : "ADVISORY "); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2108 | } else if (IS_FLOCK(fl)) { |
| 2109 | if (fl->fl_type & LOCK_MAND) { |
Pavel Emelyanov | 7f8ada9 | 2007-10-01 14:41:15 -0700 | [diff] [blame] | 2110 | seq_printf(f, "FLOCK MSNFS "); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2111 | } else { |
Pavel Emelyanov | 7f8ada9 | 2007-10-01 14:41:15 -0700 | [diff] [blame] | 2112 | seq_printf(f, "FLOCK ADVISORY "); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2113 | } |
| 2114 | } else if (IS_LEASE(fl)) { |
Pavel Emelyanov | 7f8ada9 | 2007-10-01 14:41:15 -0700 | [diff] [blame] | 2115 | seq_printf(f, "LEASE "); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2116 | if (fl->fl_type & F_INPROGRESS) |
Pavel Emelyanov | 7f8ada9 | 2007-10-01 14:41:15 -0700 | [diff] [blame] | 2117 | seq_printf(f, "BREAKING "); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2118 | else if (fl->fl_file) |
Pavel Emelyanov | 7f8ada9 | 2007-10-01 14:41:15 -0700 | [diff] [blame] | 2119 | seq_printf(f, "ACTIVE "); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2120 | else |
Pavel Emelyanov | 7f8ada9 | 2007-10-01 14:41:15 -0700 | [diff] [blame] | 2121 | seq_printf(f, "BREAKER "); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2122 | } else { |
Pavel Emelyanov | 7f8ada9 | 2007-10-01 14:41:15 -0700 | [diff] [blame] | 2123 | seq_printf(f, "UNKNOWN UNKNOWN "); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2124 | } |
| 2125 | if (fl->fl_type & LOCK_MAND) { |
Pavel Emelyanov | 7f8ada9 | 2007-10-01 14:41:15 -0700 | [diff] [blame] | 2126 | seq_printf(f, "%s ", |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2127 | (fl->fl_type & LOCK_READ) |
| 2128 | ? (fl->fl_type & LOCK_WRITE) ? "RW " : "READ " |
| 2129 | : (fl->fl_type & LOCK_WRITE) ? "WRITE" : "NONE "); |
| 2130 | } else { |
Pavel Emelyanov | 7f8ada9 | 2007-10-01 14:41:15 -0700 | [diff] [blame] | 2131 | seq_printf(f, "%s ", |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2132 | (fl->fl_type & F_INPROGRESS) |
| 2133 | ? (fl->fl_type & F_UNLCK) ? "UNLCK" : "READ " |
| 2134 | : (fl->fl_type & F_WRLCK) ? "WRITE" : "READ "); |
| 2135 | } |
| 2136 | if (inode) { |
| 2137 | #ifdef WE_CAN_BREAK_LSLK_NOW |
Vitaliy Gusev | ab1f161 | 2008-01-17 00:07:08 +0000 | [diff] [blame] | 2138 | seq_printf(f, "%d %s:%ld ", fl_pid, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2139 | inode->i_sb->s_id, inode->i_ino); |
| 2140 | #else |
| 2141 | /* userspace relies on this representation of dev_t ;-( */ |
Vitaliy Gusev | ab1f161 | 2008-01-17 00:07:08 +0000 | [diff] [blame] | 2142 | seq_printf(f, "%d %02x:%02x:%ld ", fl_pid, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2143 | MAJOR(inode->i_sb->s_dev), |
| 2144 | MINOR(inode->i_sb->s_dev), inode->i_ino); |
| 2145 | #endif |
| 2146 | } else { |
Vitaliy Gusev | ab1f161 | 2008-01-17 00:07:08 +0000 | [diff] [blame] | 2147 | seq_printf(f, "%d <none>:0 ", fl_pid); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2148 | } |
| 2149 | if (IS_POSIX(fl)) { |
| 2150 | if (fl->fl_end == OFFSET_MAX) |
Pavel Emelyanov | 7f8ada9 | 2007-10-01 14:41:15 -0700 | [diff] [blame] | 2151 | seq_printf(f, "%Ld EOF\n", fl->fl_start); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2152 | else |
Pavel Emelyanov | 7f8ada9 | 2007-10-01 14:41:15 -0700 | [diff] [blame] | 2153 | seq_printf(f, "%Ld %Ld\n", fl->fl_start, fl->fl_end); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2154 | } else { |
Pavel Emelyanov | 7f8ada9 | 2007-10-01 14:41:15 -0700 | [diff] [blame] | 2155 | seq_printf(f, "0 EOF\n"); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2156 | } |
| 2157 | } |
| 2158 | |
Pavel Emelyanov | 7f8ada9 | 2007-10-01 14:41:15 -0700 | [diff] [blame] | 2159 | static int locks_show(struct seq_file *f, void *v) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2160 | { |
Pavel Emelyanov | 7f8ada9 | 2007-10-01 14:41:15 -0700 | [diff] [blame] | 2161 | struct file_lock *fl, *bfl; |
| 2162 | |
| 2163 | fl = list_entry(v, struct file_lock, fl_link); |
| 2164 | |
| 2165 | lock_get_status(f, fl, (long)f->private, ""); |
| 2166 | |
| 2167 | list_for_each_entry(bfl, &fl->fl_block, fl_block) |
| 2168 | lock_get_status(f, bfl, (long)f->private, " ->"); |
| 2169 | |
| 2170 | f->private++; |
| 2171 | return 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2172 | } |
| 2173 | |
Pavel Emelyanov | 7f8ada9 | 2007-10-01 14:41:15 -0700 | [diff] [blame] | 2174 | static void *locks_start(struct seq_file *f, loff_t *pos) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2175 | { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2176 | lock_kernel(); |
Pavel Emelyanov | 7f8ada9 | 2007-10-01 14:41:15 -0700 | [diff] [blame] | 2177 | f->private = (void *)1; |
| 2178 | return seq_list_start(&file_lock_list, *pos); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2179 | } |
| 2180 | |
Pavel Emelyanov | 7f8ada9 | 2007-10-01 14:41:15 -0700 | [diff] [blame] | 2181 | static void *locks_next(struct seq_file *f, void *v, loff_t *pos) |
| 2182 | { |
| 2183 | return seq_list_next(v, &file_lock_list, pos); |
| 2184 | } |
| 2185 | |
| 2186 | static void locks_stop(struct seq_file *f, void *v) |
| 2187 | { |
| 2188 | unlock_kernel(); |
| 2189 | } |
| 2190 | |
| 2191 | struct seq_operations locks_seq_operations = { |
| 2192 | .start = locks_start, |
| 2193 | .next = locks_next, |
| 2194 | .stop = locks_stop, |
| 2195 | .show = locks_show, |
| 2196 | }; |
| 2197 | #endif |
| 2198 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2199 | /** |
| 2200 | * lock_may_read - checks that the region is free of locks |
| 2201 | * @inode: the inode that is being read |
| 2202 | * @start: the first byte to read |
| 2203 | * @len: the number of bytes to read |
| 2204 | * |
| 2205 | * Emulates Windows locking requirements. Whole-file |
| 2206 | * mandatory locks (share modes) can prohibit a read and |
| 2207 | * byte-range POSIX locks can prohibit a read if they overlap. |
| 2208 | * |
| 2209 | * N.B. this function is only ever called |
| 2210 | * from knfsd and ownership of locks is never checked. |
| 2211 | */ |
| 2212 | int lock_may_read(struct inode *inode, loff_t start, unsigned long len) |
| 2213 | { |
| 2214 | struct file_lock *fl; |
| 2215 | int result = 1; |
| 2216 | lock_kernel(); |
| 2217 | for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) { |
| 2218 | if (IS_POSIX(fl)) { |
| 2219 | if (fl->fl_type == F_RDLCK) |
| 2220 | continue; |
| 2221 | if ((fl->fl_end < start) || (fl->fl_start > (start + len))) |
| 2222 | continue; |
| 2223 | } else if (IS_FLOCK(fl)) { |
| 2224 | if (!(fl->fl_type & LOCK_MAND)) |
| 2225 | continue; |
| 2226 | if (fl->fl_type & LOCK_READ) |
| 2227 | continue; |
| 2228 | } else |
| 2229 | continue; |
| 2230 | result = 0; |
| 2231 | break; |
| 2232 | } |
| 2233 | unlock_kernel(); |
| 2234 | return result; |
| 2235 | } |
| 2236 | |
| 2237 | EXPORT_SYMBOL(lock_may_read); |
| 2238 | |
| 2239 | /** |
| 2240 | * lock_may_write - checks that the region is free of locks |
| 2241 | * @inode: the inode that is being written |
| 2242 | * @start: the first byte to write |
| 2243 | * @len: the number of bytes to write |
| 2244 | * |
| 2245 | * Emulates Windows locking requirements. Whole-file |
| 2246 | * mandatory locks (share modes) can prohibit a write and |
| 2247 | * byte-range POSIX locks can prohibit a write if they overlap. |
| 2248 | * |
| 2249 | * N.B. this function is only ever called |
| 2250 | * from knfsd and ownership of locks is never checked. |
| 2251 | */ |
| 2252 | int lock_may_write(struct inode *inode, loff_t start, unsigned long len) |
| 2253 | { |
| 2254 | struct file_lock *fl; |
| 2255 | int result = 1; |
| 2256 | lock_kernel(); |
| 2257 | for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) { |
| 2258 | if (IS_POSIX(fl)) { |
| 2259 | if ((fl->fl_end < start) || (fl->fl_start > (start + len))) |
| 2260 | continue; |
| 2261 | } else if (IS_FLOCK(fl)) { |
| 2262 | if (!(fl->fl_type & LOCK_MAND)) |
| 2263 | continue; |
| 2264 | if (fl->fl_type & LOCK_WRITE) |
| 2265 | continue; |
| 2266 | } else |
| 2267 | continue; |
| 2268 | result = 0; |
| 2269 | break; |
| 2270 | } |
| 2271 | unlock_kernel(); |
| 2272 | return result; |
| 2273 | } |
| 2274 | |
| 2275 | EXPORT_SYMBOL(lock_may_write); |
| 2276 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2277 | static int __init filelock_init(void) |
| 2278 | { |
| 2279 | filelock_cache = kmem_cache_create("file_lock_cache", |
| 2280 | sizeof(struct file_lock), 0, SLAB_PANIC, |
Paul Mundt | 20c2df8 | 2007-07-20 10:11:58 +0900 | [diff] [blame] | 2281 | init_once); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2282 | return 0; |
| 2283 | } |
| 2284 | |
| 2285 | core_initcall(filelock_init); |