]>
Commit | Line | Data |
---|---|---|
b3b94faa DT |
1 | /* |
2 | * Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved. | |
3 | * Copyright (C) 2004-2005 Red Hat, Inc. All rights reserved. | |
4 | * | |
5 | * This copyrighted material is made available to anyone wishing to use, | |
6 | * modify, copy, or redistribute it subject to the terms and conditions | |
7 | * of the GNU General Public License v.2. | |
8 | */ | |
9 | ||
10 | #include <linux/sched.h> | |
11 | #include <linux/slab.h> | |
12 | #include <linux/spinlock.h> | |
13 | #include <linux/completion.h> | |
14 | #include <linux/buffer_head.h> | |
15 | #include <linux/pagemap.h> | |
16 | #include <linux/uio.h> | |
17 | #include <linux/blkdev.h> | |
18 | #include <linux/mm.h> | |
19 | #include <linux/smp_lock.h> | |
20 | #include <linux/gfs2_ioctl.h> | |
18ec7d5c | 21 | #include <linux/fs.h> |
b3b94faa DT |
22 | #include <asm/semaphore.h> |
23 | #include <asm/uaccess.h> | |
24 | ||
25 | #include "gfs2.h" | |
26 | #include "bmap.h" | |
27 | #include "dir.h" | |
28 | #include "glock.h" | |
29 | #include "glops.h" | |
30 | #include "inode.h" | |
b3b94faa DT |
31 | #include "lm.h" |
32 | #include "log.h" | |
33 | #include "meta_io.h" | |
34 | #include "ops_file.h" | |
35 | #include "ops_vm.h" | |
36 | #include "quota.h" | |
37 | #include "rgrp.h" | |
38 | #include "trans.h" | |
39 | ||
40 | /* "bad" is for NFS support */ | |
41 | struct filldir_bad_entry { | |
42 | char *fbe_name; | |
43 | unsigned int fbe_length; | |
44 | uint64_t fbe_offset; | |
45 | struct gfs2_inum fbe_inum; | |
46 | unsigned int fbe_type; | |
47 | }; | |
48 | ||
49 | struct filldir_bad { | |
50 | struct gfs2_sbd *fdb_sbd; | |
51 | ||
52 | struct filldir_bad_entry *fdb_entry; | |
53 | unsigned int fdb_entry_num; | |
54 | unsigned int fdb_entry_off; | |
55 | ||
56 | char *fdb_name; | |
57 | unsigned int fdb_name_size; | |
58 | unsigned int fdb_name_off; | |
59 | }; | |
60 | ||
61 | /* For regular, non-NFS */ | |
62 | struct filldir_reg { | |
63 | struct gfs2_sbd *fdr_sbd; | |
64 | int fdr_prefetch; | |
65 | ||
66 | filldir_t fdr_filldir; | |
67 | void *fdr_opaque; | |
68 | }; | |
69 | ||
18ec7d5c SW |
70 | static int gfs2_read_actor(read_descriptor_t *desc, struct page *page, |
71 | unsigned long offset, unsigned long size) | |
72 | { | |
73 | char *kaddr; | |
74 | unsigned long count = desc->count; | |
75 | ||
76 | if (size > count) | |
77 | size = count; | |
78 | ||
79 | kaddr = kmap(page); | |
80 | memcpy(desc->arg.buf, kaddr + offset, size); | |
81 | kunmap(page); | |
82 | ||
83 | desc->count = count - size; | |
84 | desc->written += size; | |
85 | desc->arg.buf += size; | |
86 | return size; | |
87 | } | |
88 | ||
89 | int gfs2_internal_read(struct gfs2_inode *ip, struct file_ra_state *ra_state, | |
90 | char *buf, loff_t *pos, unsigned size) | |
91 | { | |
92 | struct inode *inode = ip->i_vnode; | |
93 | read_descriptor_t desc; | |
94 | desc.written = 0; | |
95 | desc.arg.buf = buf; | |
96 | desc.count = size; | |
97 | desc.error = 0; | |
98 | do_generic_mapping_read(inode->i_mapping, ra_state, NULL, pos, &desc, gfs2_read_actor); | |
99 | return desc.written ? desc.written : desc.error; | |
100 | } | |
b3b94faa DT |
101 | |
102 | /** | |
103 | * gfs2_llseek - seek to a location in a file | |
104 | * @file: the file | |
105 | * @offset: the offset | |
106 | * @origin: Where to seek from (SEEK_SET, SEEK_CUR, or SEEK_END) | |
107 | * | |
108 | * SEEK_END requires the glock for the file because it references the | |
109 | * file's size. | |
110 | * | |
111 | * Returns: The new offset, or errno | |
112 | */ | |
113 | ||
114 | static loff_t gfs2_llseek(struct file *file, loff_t offset, int origin) | |
115 | { | |
116 | struct gfs2_inode *ip = get_v2ip(file->f_mapping->host); | |
117 | struct gfs2_holder i_gh; | |
118 | loff_t error; | |
119 | ||
120 | atomic_inc(&ip->i_sbd->sd_ops_file); | |
121 | ||
122 | if (origin == 2) { | |
123 | error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED, LM_FLAG_ANY, | |
124 | &i_gh); | |
125 | if (!error) { | |
126 | error = remote_llseek(file, offset, origin); | |
127 | gfs2_glock_dq_uninit(&i_gh); | |
128 | } | |
129 | } else | |
130 | error = remote_llseek(file, offset, origin); | |
131 | ||
132 | return error; | |
133 | } | |
134 | ||
b3b94faa | 135 | |
18ec7d5c SW |
136 | static ssize_t gfs2_direct_IO_read(struct kiocb *iocb, const struct iovec *iov, |
137 | loff_t offset, unsigned long nr_segs) | |
b3b94faa | 138 | { |
18ec7d5c SW |
139 | struct file *file = iocb->ki_filp; |
140 | struct address_space *mapping = file->f_mapping; | |
141 | ssize_t retval; | |
b3b94faa | 142 | |
18ec7d5c SW |
143 | retval = filemap_write_and_wait(mapping); |
144 | if (retval == 0) { | |
145 | retval = mapping->a_ops->direct_IO(READ, iocb, iov, offset, | |
146 | nr_segs); | |
b3b94faa | 147 | } |
18ec7d5c | 148 | return retval; |
b3b94faa DT |
149 | } |
150 | ||
151 | /** | |
18ec7d5c SW |
152 | * __gfs2_file_aio_read - The main GFS2 read function |
153 | * | |
154 | * N.B. This is almost, but not quite the same as __generic_file_aio_read() | |
155 | * the important subtle different being that inode->i_size isn't valid | |
156 | * unless we are holding a lock, and we do this _only_ on the O_DIRECT | |
157 | * path since otherwise locking is done entirely at the page cache | |
158 | * layer. | |
b3b94faa | 159 | */ |
18ec7d5c SW |
160 | static ssize_t __gfs2_file_aio_read(struct kiocb *iocb, |
161 | const struct iovec *iov, | |
162 | unsigned long nr_segs, loff_t *ppos) | |
b3b94faa | 163 | { |
18ec7d5c SW |
164 | struct file *filp = iocb->ki_filp; |
165 | struct gfs2_inode *ip = get_v2ip(filp->f_mapping->host); | |
b3b94faa | 166 | struct gfs2_holder gh; |
18ec7d5c SW |
167 | ssize_t retval; |
168 | unsigned long seg; | |
169 | size_t count; | |
170 | ||
171 | count = 0; | |
172 | for (seg = 0; seg < nr_segs; seg++) { | |
173 | const struct iovec *iv = &iov[seg]; | |
174 | ||
175 | /* | |
176 | * If any segment has a negative length, or the cumulative | |
177 | * length ever wraps negative then return -EINVAL. | |
178 | */ | |
179 | count += iv->iov_len; | |
180 | if (unlikely((ssize_t)(count|iv->iov_len) < 0)) | |
b3b94faa | 181 | return -EINVAL; |
18ec7d5c SW |
182 | if (access_ok(VERIFY_WRITE, iv->iov_base, iv->iov_len)) |
183 | continue; | |
184 | if (seg == 0) | |
b3b94faa | 185 | return -EFAULT; |
18ec7d5c SW |
186 | nr_segs = seg; |
187 | count -= iv->iov_len; /* This segment is no good */ | |
188 | break; | |
189 | } | |
190 | ||
191 | /* coalesce the iovecs and go direct-to-BIO for O_DIRECT */ | |
192 | if (filp->f_flags & O_DIRECT) { | |
193 | loff_t pos = *ppos, size; | |
194 | struct address_space *mapping; | |
195 | struct inode *inode; | |
196 | ||
197 | mapping = filp->f_mapping; | |
198 | inode = mapping->host; | |
199 | retval = 0; | |
200 | if (!count) | |
201 | goto out; /* skip atime */ | |
202 | ||
203 | gfs2_holder_init(ip->i_gl, LM_ST_SHARED, GL_ATIME, &gh); | |
204 | retval = gfs2_glock_nq_m_atime(1, &gh); | |
205 | if (retval) | |
206 | goto out; | |
b3b94faa | 207 | |
18ec7d5c SW |
208 | size = i_size_read(inode); |
209 | if (pos < size) { | |
210 | retval = gfs2_direct_IO_read(iocb, iov, pos, nr_segs); | |
211 | if (retval > 0 && !is_sync_kiocb(iocb)) | |
212 | retval = -EIOCBQUEUED; | |
213 | if (retval > 0) | |
214 | *ppos = pos + retval; | |
b3b94faa | 215 | } |
18ec7d5c SW |
216 | file_accessed(filp); |
217 | gfs2_glock_dq_m(1, &gh); | |
218 | gfs2_holder_uninit(&gh); | |
b3b94faa | 219 | goto out; |
18ec7d5c | 220 | } |
b3b94faa | 221 | |
18ec7d5c SW |
222 | retval = 0; |
223 | if (count) { | |
224 | for (seg = 0; seg < nr_segs; seg++) { | |
225 | read_descriptor_t desc; | |
226 | ||
227 | desc.written = 0; | |
228 | desc.arg.buf = iov[seg].iov_base; | |
229 | desc.count = iov[seg].iov_len; | |
230 | if (desc.count == 0) | |
231 | continue; | |
232 | desc.error = 0; | |
233 | do_generic_file_read(filp,ppos,&desc,file_read_actor); | |
234 | retval += desc.written; | |
235 | if (desc.error) { | |
236 | retval = retval ?: desc.error; | |
237 | break; | |
238 | } | |
239 | } | |
240 | } | |
241 | out: | |
242 | return retval; | |
b3b94faa DT |
243 | } |
244 | ||
245 | /** | |
246 | * gfs2_read - Read bytes from a file | |
247 | * @file: The file to read from | |
248 | * @buf: The buffer to copy into | |
249 | * @size: The amount of data requested | |
250 | * @offset: The current file offset | |
251 | * | |
252 | * Outputs: Offset - updated according to number of bytes read | |
253 | * | |
254 | * Returns: The number of bytes read, errno on failure | |
255 | */ | |
256 | ||
18ec7d5c | 257 | static ssize_t gfs2_read(struct file *filp, char __user *buf, size_t size, |
b3b94faa DT |
258 | loff_t *offset) |
259 | { | |
b3b94faa | 260 | struct iovec local_iov = { .iov_base = buf, .iov_len = size }; |
18ec7d5c SW |
261 | struct kiocb kiocb; |
262 | ssize_t ret; | |
b3b94faa | 263 | |
18ec7d5c | 264 | atomic_inc(&get_v2sdp(filp->f_mapping->host->i_sb)->sd_ops_file); |
b3b94faa | 265 | |
18ec7d5c SW |
266 | init_sync_kiocb(&kiocb, filp); |
267 | ret = __gfs2_file_aio_read(&kiocb, &local_iov, 1, offset); | |
268 | if (-EIOCBQUEUED == ret) | |
269 | ret = wait_on_sync_kiocb(&kiocb); | |
270 | return ret; | |
b3b94faa DT |
271 | } |
272 | ||
18ec7d5c SW |
273 | static ssize_t gfs2_file_readv(struct file *filp, const struct iovec *iov, |
274 | unsigned long nr_segs, loff_t *ppos) | |
b3b94faa | 275 | { |
18ec7d5c SW |
276 | struct kiocb kiocb; |
277 | ssize_t ret; | |
b3b94faa | 278 | |
18ec7d5c | 279 | atomic_inc(&get_v2sdp(filp->f_mapping->host->i_sb)->sd_ops_file); |
b3b94faa | 280 | |
18ec7d5c SW |
281 | init_sync_kiocb(&kiocb, filp); |
282 | ret = __gfs2_file_aio_read(&kiocb, iov, nr_segs, ppos); | |
283 | if (-EIOCBQUEUED == ret) | |
284 | ret = wait_on_sync_kiocb(&kiocb); | |
285 | return ret; | |
b3b94faa DT |
286 | } |
287 | ||
18ec7d5c SW |
288 | static ssize_t gfs2_file_aio_read(struct kiocb *iocb, char __user *buf, |
289 | size_t count, loff_t pos) | |
b3b94faa | 290 | { |
18ec7d5c SW |
291 | struct file *filp = iocb->ki_filp; |
292 | struct iovec local_iov = { .iov_base = buf, .iov_len = count }; | |
b3b94faa | 293 | |
18ec7d5c | 294 | atomic_inc(&get_v2sdp(filp->f_mapping->host->i_sb)->sd_ops_file); |
b3b94faa | 295 | |
18ec7d5c SW |
296 | BUG_ON(iocb->ki_pos != pos); |
297 | return __gfs2_file_aio_read(iocb, &local_iov, 1, &iocb->ki_pos); | |
b3b94faa DT |
298 | } |
299 | ||
b3b94faa DT |
300 | |
301 | /** | |
302 | * filldir_reg_func - Report a directory entry to the caller of gfs2_dir_read() | |
303 | * @opaque: opaque data used by the function | |
304 | * @name: the name of the directory entry | |
305 | * @length: the length of the name | |
306 | * @offset: the entry's offset in the directory | |
307 | * @inum: the inode number the entry points to | |
308 | * @type: the type of inode the entry points to | |
309 | * | |
310 | * Returns: 0 on success, 1 if buffer full | |
311 | */ | |
312 | ||
313 | static int filldir_reg_func(void *opaque, const char *name, unsigned int length, | |
314 | uint64_t offset, struct gfs2_inum *inum, | |
315 | unsigned int type) | |
316 | { | |
317 | struct filldir_reg *fdr = (struct filldir_reg *)opaque; | |
318 | struct gfs2_sbd *sdp = fdr->fdr_sbd; | |
319 | int error; | |
320 | ||
321 | error = fdr->fdr_filldir(fdr->fdr_opaque, name, length, offset, | |
322 | inum->no_formal_ino, type); | |
323 | if (error) | |
324 | return 1; | |
325 | ||
326 | if (fdr->fdr_prefetch && !(length == 1 && *name == '.')) { | |
327 | gfs2_glock_prefetch_num(sdp, | |
328 | inum->no_addr, &gfs2_inode_glops, | |
329 | LM_ST_SHARED, LM_FLAG_TRY | LM_FLAG_ANY); | |
330 | gfs2_glock_prefetch_num(sdp, | |
331 | inum->no_addr, &gfs2_iopen_glops, | |
332 | LM_ST_SHARED, LM_FLAG_TRY); | |
333 | } | |
334 | ||
335 | return 0; | |
336 | } | |
337 | ||
338 | /** | |
339 | * readdir_reg - Read directory entries from a directory | |
340 | * @file: The directory to read from | |
341 | * @dirent: Buffer for dirents | |
342 | * @filldir: Function used to do the copying | |
343 | * | |
344 | * Returns: errno | |
345 | */ | |
346 | ||
347 | static int readdir_reg(struct file *file, void *dirent, filldir_t filldir) | |
348 | { | |
349 | struct gfs2_inode *dip = get_v2ip(file->f_mapping->host); | |
350 | struct filldir_reg fdr; | |
351 | struct gfs2_holder d_gh; | |
352 | uint64_t offset = file->f_pos; | |
353 | int error; | |
354 | ||
355 | fdr.fdr_sbd = dip->i_sbd; | |
356 | fdr.fdr_prefetch = 1; | |
357 | fdr.fdr_filldir = filldir; | |
358 | fdr.fdr_opaque = dirent; | |
359 | ||
360 | gfs2_holder_init(dip->i_gl, LM_ST_SHARED, GL_ATIME, &d_gh); | |
361 | error = gfs2_glock_nq_atime(&d_gh); | |
362 | if (error) { | |
363 | gfs2_holder_uninit(&d_gh); | |
364 | return error; | |
365 | } | |
366 | ||
367 | error = gfs2_dir_read(dip, &offset, &fdr, filldir_reg_func); | |
368 | ||
369 | gfs2_glock_dq_uninit(&d_gh); | |
370 | ||
371 | file->f_pos = offset; | |
372 | ||
373 | return error; | |
374 | } | |
375 | ||
376 | /** | |
377 | * filldir_bad_func - Report a directory entry to the caller of gfs2_dir_read() | |
378 | * @opaque: opaque data used by the function | |
379 | * @name: the name of the directory entry | |
380 | * @length: the length of the name | |
381 | * @offset: the entry's offset in the directory | |
382 | * @inum: the inode number the entry points to | |
383 | * @type: the type of inode the entry points to | |
384 | * | |
385 | * For supporting NFS. | |
386 | * | |
387 | * Returns: 0 on success, 1 if buffer full | |
388 | */ | |
389 | ||
390 | static int filldir_bad_func(void *opaque, const char *name, unsigned int length, | |
391 | uint64_t offset, struct gfs2_inum *inum, | |
392 | unsigned int type) | |
393 | { | |
394 | struct filldir_bad *fdb = (struct filldir_bad *)opaque; | |
395 | struct gfs2_sbd *sdp = fdb->fdb_sbd; | |
396 | struct filldir_bad_entry *fbe; | |
397 | ||
398 | if (fdb->fdb_entry_off == fdb->fdb_entry_num || | |
399 | fdb->fdb_name_off + length > fdb->fdb_name_size) | |
400 | return 1; | |
401 | ||
402 | fbe = &fdb->fdb_entry[fdb->fdb_entry_off]; | |
403 | fbe->fbe_name = fdb->fdb_name + fdb->fdb_name_off; | |
404 | memcpy(fbe->fbe_name, name, length); | |
405 | fbe->fbe_length = length; | |
406 | fbe->fbe_offset = offset; | |
407 | fbe->fbe_inum = *inum; | |
408 | fbe->fbe_type = type; | |
409 | ||
410 | fdb->fdb_entry_off++; | |
411 | fdb->fdb_name_off += length; | |
412 | ||
413 | if (!(length == 1 && *name == '.')) { | |
414 | gfs2_glock_prefetch_num(sdp, | |
415 | inum->no_addr, &gfs2_inode_glops, | |
416 | LM_ST_SHARED, LM_FLAG_TRY | LM_FLAG_ANY); | |
417 | gfs2_glock_prefetch_num(sdp, | |
418 | inum->no_addr, &gfs2_iopen_glops, | |
419 | LM_ST_SHARED, LM_FLAG_TRY); | |
420 | } | |
421 | ||
422 | return 0; | |
423 | } | |
424 | ||
425 | /** | |
426 | * readdir_bad - Read directory entries from a directory | |
427 | * @file: The directory to read from | |
428 | * @dirent: Buffer for dirents | |
429 | * @filldir: Function used to do the copying | |
430 | * | |
431 | * For supporting NFS. | |
432 | * | |
433 | * Returns: errno | |
434 | */ | |
435 | ||
436 | static int readdir_bad(struct file *file, void *dirent, filldir_t filldir) | |
437 | { | |
438 | struct gfs2_inode *dip = get_v2ip(file->f_mapping->host); | |
439 | struct gfs2_sbd *sdp = dip->i_sbd; | |
440 | struct filldir_reg fdr; | |
441 | unsigned int entries, size; | |
442 | struct filldir_bad *fdb; | |
443 | struct gfs2_holder d_gh; | |
444 | uint64_t offset = file->f_pos; | |
445 | unsigned int x; | |
446 | struct filldir_bad_entry *fbe; | |
447 | int error; | |
448 | ||
449 | entries = gfs2_tune_get(sdp, gt_entries_per_readdir); | |
450 | size = sizeof(struct filldir_bad) + | |
451 | entries * (sizeof(struct filldir_bad_entry) + GFS2_FAST_NAME_SIZE); | |
452 | ||
453 | fdb = kzalloc(size, GFP_KERNEL); | |
454 | if (!fdb) | |
455 | return -ENOMEM; | |
456 | ||
457 | fdb->fdb_sbd = sdp; | |
458 | fdb->fdb_entry = (struct filldir_bad_entry *)(fdb + 1); | |
459 | fdb->fdb_entry_num = entries; | |
460 | fdb->fdb_name = ((char *)fdb) + sizeof(struct filldir_bad) + | |
461 | entries * sizeof(struct filldir_bad_entry); | |
462 | fdb->fdb_name_size = entries * GFS2_FAST_NAME_SIZE; | |
463 | ||
464 | gfs2_holder_init(dip->i_gl, LM_ST_SHARED, GL_ATIME, &d_gh); | |
465 | error = gfs2_glock_nq_atime(&d_gh); | |
466 | if (error) { | |
467 | gfs2_holder_uninit(&d_gh); | |
468 | goto out; | |
469 | } | |
470 | ||
471 | error = gfs2_dir_read(dip, &offset, fdb, filldir_bad_func); | |
472 | ||
473 | gfs2_glock_dq_uninit(&d_gh); | |
474 | ||
475 | fdr.fdr_sbd = sdp; | |
476 | fdr.fdr_prefetch = 0; | |
477 | fdr.fdr_filldir = filldir; | |
478 | fdr.fdr_opaque = dirent; | |
479 | ||
480 | for (x = 0; x < fdb->fdb_entry_off; x++) { | |
481 | fbe = &fdb->fdb_entry[x]; | |
482 | ||
483 | error = filldir_reg_func(&fdr, | |
484 | fbe->fbe_name, fbe->fbe_length, | |
485 | fbe->fbe_offset, | |
486 | &fbe->fbe_inum, fbe->fbe_type); | |
487 | if (error) { | |
488 | file->f_pos = fbe->fbe_offset; | |
489 | error = 0; | |
490 | goto out; | |
491 | } | |
492 | } | |
493 | ||
494 | file->f_pos = offset; | |
495 | ||
496 | out: | |
497 | kfree(fdb); | |
498 | ||
499 | return error; | |
500 | } | |
501 | ||
502 | /** | |
503 | * gfs2_readdir - Read directory entries from a directory | |
504 | * @file: The directory to read from | |
505 | * @dirent: Buffer for dirents | |
506 | * @filldir: Function used to do the copying | |
507 | * | |
508 | * Returns: errno | |
509 | */ | |
510 | ||
511 | static int gfs2_readdir(struct file *file, void *dirent, filldir_t filldir) | |
512 | { | |
513 | int error; | |
514 | ||
515 | atomic_inc(&get_v2sdp(file->f_mapping->host->i_sb)->sd_ops_file); | |
516 | ||
517 | if (strcmp(current->comm, "nfsd") != 0) | |
518 | error = readdir_reg(file, dirent, filldir); | |
519 | else | |
520 | error = readdir_bad(file, dirent, filldir); | |
521 | ||
522 | return error; | |
523 | } | |
524 | ||
525 | static int gfs2_ioctl_flags(struct gfs2_inode *ip, unsigned int cmd, unsigned long arg) | |
526 | { | |
527 | unsigned int lmode = (cmd == GFS2_IOCTL_SETFLAGS) ? LM_ST_EXCLUSIVE : LM_ST_SHARED; | |
528 | struct buffer_head *dibh; | |
529 | struct gfs2_holder i_gh; | |
530 | int error; | |
531 | __u32 flags = 0, change; | |
532 | ||
533 | if (cmd == GFS2_IOCTL_SETFLAGS) { | |
534 | error = get_user(flags, (__u32 __user *)arg); | |
535 | if (error) | |
536 | return -EFAULT; | |
537 | } | |
538 | ||
539 | error = gfs2_glock_nq_init(ip->i_gl, lmode, 0, &i_gh); | |
540 | if (error) | |
541 | return error; | |
542 | ||
543 | if (cmd == GFS2_IOCTL_SETFLAGS) { | |
544 | change = flags ^ ip->i_di.di_flags; | |
545 | error = -EPERM; | |
546 | if (change & (GFS2_DIF_IMMUTABLE|GFS2_DIF_APPENDONLY)) { | |
547 | if (!capable(CAP_LINUX_IMMUTABLE)) | |
548 | goto out; | |
549 | } | |
550 | error = -EINVAL; | |
551 | if (flags & (GFS2_DIF_JDATA|GFS2_DIF_DIRECTIO)) { | |
552 | if (!S_ISREG(ip->i_di.di_mode)) | |
553 | goto out; | |
b3b94faa DT |
554 | } |
555 | if (flags & (GFS2_DIF_INHERIT_JDATA|GFS2_DIF_INHERIT_DIRECTIO)) { | |
556 | if (!S_ISDIR(ip->i_di.di_mode)) | |
557 | goto out; | |
558 | } | |
559 | ||
560 | error = gfs2_trans_begin(ip->i_sbd, RES_DINODE, 0); | |
561 | if (error) | |
562 | goto out; | |
563 | ||
564 | error = gfs2_meta_inode_buffer(ip, &dibh); | |
565 | if (error) | |
566 | goto out_trans_end; | |
567 | ||
568 | ip->i_di.di_flags = flags; | |
569 | ||
d4e9c4c3 | 570 | gfs2_trans_add_bh(ip->i_gl, dibh, 1); |
b3b94faa DT |
571 | gfs2_dinode_out(&ip->i_di, dibh->b_data); |
572 | ||
573 | brelse(dibh); | |
574 | ||
575 | out_trans_end: | |
576 | gfs2_trans_end(ip->i_sbd); | |
577 | } else { | |
578 | flags = ip->i_di.di_flags; | |
579 | } | |
580 | out: | |
581 | gfs2_glock_dq_uninit(&i_gh); | |
582 | if (cmd == GFS2_IOCTL_GETFLAGS) { | |
583 | if (put_user(flags, (__u32 __user *)arg)) | |
584 | return -EFAULT; | |
585 | } | |
586 | return error; | |
587 | } | |
588 | ||
589 | /** | |
590 | * gfs2_ioctl - do an ioctl on a file | |
591 | * @inode: the inode | |
592 | * @file: the file pointer | |
593 | * @cmd: the ioctl command | |
594 | * @arg: the argument | |
595 | * | |
596 | * Returns: errno | |
597 | */ | |
598 | ||
599 | static int gfs2_ioctl(struct inode *inode, struct file *file, unsigned int cmd, | |
600 | unsigned long arg) | |
601 | { | |
602 | struct gfs2_inode *ip = get_v2ip(inode); | |
603 | ||
604 | atomic_inc(&ip->i_sbd->sd_ops_file); | |
605 | ||
606 | switch (cmd) { | |
b3b94faa DT |
607 | case GFS2_IOCTL_SETFLAGS: |
608 | case GFS2_IOCTL_GETFLAGS: | |
609 | return gfs2_ioctl_flags(ip, cmd, arg); | |
b3b94faa DT |
610 | |
611 | default: | |
612 | return -ENOTTY; | |
613 | } | |
614 | } | |
615 | ||
616 | /** | |
617 | * gfs2_mmap - | |
618 | * @file: The file to map | |
619 | * @vma: The VMA which described the mapping | |
620 | * | |
621 | * Returns: 0 or error code | |
622 | */ | |
623 | ||
624 | static int gfs2_mmap(struct file *file, struct vm_area_struct *vma) | |
625 | { | |
626 | struct gfs2_inode *ip = get_v2ip(file->f_mapping->host); | |
627 | struct gfs2_holder i_gh; | |
628 | int error; | |
629 | ||
630 | atomic_inc(&ip->i_sbd->sd_ops_file); | |
631 | ||
632 | gfs2_holder_init(ip->i_gl, LM_ST_SHARED, GL_ATIME, &i_gh); | |
633 | error = gfs2_glock_nq_atime(&i_gh); | |
634 | if (error) { | |
635 | gfs2_holder_uninit(&i_gh); | |
636 | return error; | |
637 | } | |
638 | ||
18ec7d5c SW |
639 | /* This is VM_MAYWRITE instead of VM_WRITE because a call |
640 | to mprotect() can turn on VM_WRITE later. */ | |
641 | ||
642 | if ((vma->vm_flags & (VM_MAYSHARE | VM_MAYWRITE)) == | |
643 | (VM_MAYSHARE | VM_MAYWRITE)) | |
644 | vma->vm_ops = &gfs2_vm_ops_sharewrite; | |
645 | else | |
646 | vma->vm_ops = &gfs2_vm_ops_private; | |
b3b94faa DT |
647 | |
648 | gfs2_glock_dq_uninit(&i_gh); | |
649 | ||
650 | return error; | |
651 | } | |
652 | ||
653 | /** | |
654 | * gfs2_open - open a file | |
655 | * @inode: the inode to open | |
656 | * @file: the struct file for this opening | |
657 | * | |
658 | * Returns: errno | |
659 | */ | |
660 | ||
661 | static int gfs2_open(struct inode *inode, struct file *file) | |
662 | { | |
663 | struct gfs2_inode *ip = get_v2ip(inode); | |
664 | struct gfs2_holder i_gh; | |
665 | struct gfs2_file *fp; | |
666 | int error; | |
667 | ||
668 | atomic_inc(&ip->i_sbd->sd_ops_file); | |
669 | ||
670 | fp = kzalloc(sizeof(struct gfs2_file), GFP_KERNEL); | |
671 | if (!fp) | |
672 | return -ENOMEM; | |
673 | ||
674 | init_MUTEX(&fp->f_fl_mutex); | |
675 | ||
676 | fp->f_inode = ip; | |
677 | fp->f_vfile = file; | |
678 | ||
679 | gfs2_assert_warn(ip->i_sbd, !get_v2fp(file)); | |
680 | set_v2fp(file, fp); | |
681 | ||
682 | if (S_ISREG(ip->i_di.di_mode)) { | |
683 | error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED, LM_FLAG_ANY, | |
684 | &i_gh); | |
685 | if (error) | |
686 | goto fail; | |
687 | ||
688 | if (!(file->f_flags & O_LARGEFILE) && | |
689 | ip->i_di.di_size > MAX_NON_LFS) { | |
690 | error = -EFBIG; | |
691 | goto fail_gunlock; | |
692 | } | |
693 | ||
694 | /* Listen to the Direct I/O flag */ | |
695 | ||
696 | if (ip->i_di.di_flags & GFS2_DIF_DIRECTIO) | |
697 | file->f_flags |= O_DIRECT; | |
698 | ||
b3b94faa DT |
699 | gfs2_glock_dq_uninit(&i_gh); |
700 | } | |
701 | ||
702 | return 0; | |
703 | ||
704 | fail_gunlock: | |
705 | gfs2_glock_dq_uninit(&i_gh); | |
706 | ||
707 | fail: | |
708 | set_v2fp(file, NULL); | |
709 | kfree(fp); | |
710 | ||
711 | return error; | |
712 | } | |
713 | ||
714 | /** | |
715 | * gfs2_close - called to close a struct file | |
716 | * @inode: the inode the struct file belongs to | |
717 | * @file: the struct file being closed | |
718 | * | |
719 | * Returns: errno | |
720 | */ | |
721 | ||
722 | static int gfs2_close(struct inode *inode, struct file *file) | |
723 | { | |
724 | struct gfs2_sbd *sdp = get_v2sdp(inode->i_sb); | |
725 | struct gfs2_file *fp; | |
726 | ||
727 | atomic_inc(&sdp->sd_ops_file); | |
728 | ||
729 | fp = get_v2fp(file); | |
730 | set_v2fp(file, NULL); | |
731 | ||
732 | if (gfs2_assert_warn(sdp, fp)) | |
733 | return -EIO; | |
734 | ||
735 | kfree(fp); | |
736 | ||
737 | return 0; | |
738 | } | |
739 | ||
740 | /** | |
741 | * gfs2_fsync - sync the dirty data for a file (across the cluster) | |
742 | * @file: the file that points to the dentry (we ignore this) | |
743 | * @dentry: the dentry that points to the inode to sync | |
744 | * | |
745 | * Returns: errno | |
746 | */ | |
747 | ||
748 | static int gfs2_fsync(struct file *file, struct dentry *dentry, int datasync) | |
749 | { | |
750 | struct gfs2_inode *ip = get_v2ip(dentry->d_inode); | |
751 | ||
752 | atomic_inc(&ip->i_sbd->sd_ops_file); | |
753 | gfs2_log_flush_glock(ip->i_gl); | |
754 | ||
755 | return 0; | |
756 | } | |
757 | ||
758 | /** | |
759 | * gfs2_lock - acquire/release a posix lock on a file | |
760 | * @file: the file pointer | |
761 | * @cmd: either modify or retrieve lock state, possibly wait | |
762 | * @fl: type and range of lock | |
763 | * | |
764 | * Returns: errno | |
765 | */ | |
766 | ||
767 | static int gfs2_lock(struct file *file, int cmd, struct file_lock *fl) | |
768 | { | |
769 | struct gfs2_inode *ip = get_v2ip(file->f_mapping->host); | |
770 | struct gfs2_sbd *sdp = ip->i_sbd; | |
771 | struct lm_lockname name = | |
772 | { .ln_number = ip->i_num.no_addr, | |
773 | .ln_type = LM_TYPE_PLOCK }; | |
774 | ||
775 | atomic_inc(&sdp->sd_ops_file); | |
776 | ||
777 | if (!(fl->fl_flags & FL_POSIX)) | |
778 | return -ENOLCK; | |
779 | if ((ip->i_di.di_mode & (S_ISGID | S_IXGRP)) == S_ISGID) | |
780 | return -ENOLCK; | |
781 | ||
782 | if (sdp->sd_args.ar_localflocks) { | |
783 | if (IS_GETLK(cmd)) { | |
784 | struct file_lock *tmp; | |
785 | lock_kernel(); | |
786 | tmp = posix_test_lock(file, fl); | |
787 | fl->fl_type = F_UNLCK; | |
788 | if (tmp) | |
789 | memcpy(fl, tmp, sizeof(struct file_lock)); | |
790 | unlock_kernel(); | |
791 | return 0; | |
792 | } else { | |
793 | int error; | |
794 | lock_kernel(); | |
795 | error = posix_lock_file_wait(file, fl); | |
796 | unlock_kernel(); | |
797 | return error; | |
798 | } | |
799 | } | |
800 | ||
801 | if (IS_GETLK(cmd)) | |
802 | return gfs2_lm_plock_get(sdp, &name, file, fl); | |
803 | else if (fl->fl_type == F_UNLCK) | |
804 | return gfs2_lm_punlock(sdp, &name, file, fl); | |
805 | else | |
806 | return gfs2_lm_plock(sdp, &name, file, cmd, fl); | |
807 | } | |
808 | ||
809 | /** | |
810 | * gfs2_sendfile - Send bytes to a file or socket | |
811 | * @in_file: The file to read from | |
812 | * @out_file: The file to write to | |
813 | * @count: The amount of data | |
814 | * @offset: The beginning file offset | |
815 | * | |
816 | * Outputs: offset - updated according to number of bytes read | |
817 | * | |
818 | * Returns: The number of bytes sent, errno on failure | |
819 | */ | |
820 | ||
821 | static ssize_t gfs2_sendfile(struct file *in_file, loff_t *offset, size_t count, | |
822 | read_actor_t actor, void *target) | |
823 | { | |
824 | struct gfs2_inode *ip = get_v2ip(in_file->f_mapping->host); | |
b3b94faa DT |
825 | |
826 | atomic_inc(&ip->i_sbd->sd_ops_file); | |
827 | ||
18ec7d5c | 828 | return generic_file_sendfile(in_file, offset, count, actor, target); |
b3b94faa DT |
829 | } |
830 | ||
831 | static int do_flock(struct file *file, int cmd, struct file_lock *fl) | |
832 | { | |
833 | struct gfs2_file *fp = get_v2fp(file); | |
834 | struct gfs2_holder *fl_gh = &fp->f_fl_gh; | |
835 | struct gfs2_inode *ip = fp->f_inode; | |
836 | struct gfs2_glock *gl; | |
837 | unsigned int state; | |
838 | int flags; | |
839 | int error = 0; | |
840 | ||
841 | state = (fl->fl_type == F_WRLCK) ? LM_ST_EXCLUSIVE : LM_ST_SHARED; | |
842 | flags = ((IS_SETLKW(cmd)) ? 0 : LM_FLAG_TRY) | GL_EXACT | GL_NOCACHE; | |
843 | ||
844 | down(&fp->f_fl_mutex); | |
845 | ||
846 | gl = fl_gh->gh_gl; | |
847 | if (gl) { | |
848 | if (fl_gh->gh_state == state) | |
849 | goto out; | |
850 | gfs2_glock_hold(gl); | |
851 | flock_lock_file_wait(file, | |
852 | &(struct file_lock){.fl_type = F_UNLCK}); | |
853 | gfs2_glock_dq_uninit(fl_gh); | |
854 | } else { | |
855 | error = gfs2_glock_get(ip->i_sbd, | |
856 | ip->i_num.no_addr, &gfs2_flock_glops, | |
857 | CREATE, &gl); | |
858 | if (error) | |
859 | goto out; | |
860 | } | |
861 | ||
862 | gfs2_holder_init(gl, state, flags, fl_gh); | |
863 | gfs2_glock_put(gl); | |
864 | ||
865 | error = gfs2_glock_nq(fl_gh); | |
866 | if (error) { | |
867 | gfs2_holder_uninit(fl_gh); | |
868 | if (error == GLR_TRYFAILED) | |
869 | error = -EAGAIN; | |
870 | } else { | |
871 | error = flock_lock_file_wait(file, fl); | |
872 | gfs2_assert_warn(ip->i_sbd, !error); | |
873 | } | |
874 | ||
875 | out: | |
876 | up(&fp->f_fl_mutex); | |
877 | ||
878 | return error; | |
879 | } | |
880 | ||
881 | static void do_unflock(struct file *file, struct file_lock *fl) | |
882 | { | |
883 | struct gfs2_file *fp = get_v2fp(file); | |
884 | struct gfs2_holder *fl_gh = &fp->f_fl_gh; | |
885 | ||
886 | down(&fp->f_fl_mutex); | |
887 | flock_lock_file_wait(file, fl); | |
888 | if (fl_gh->gh_gl) | |
889 | gfs2_glock_dq_uninit(fl_gh); | |
890 | up(&fp->f_fl_mutex); | |
891 | } | |
892 | ||
893 | /** | |
894 | * gfs2_flock - acquire/release a flock lock on a file | |
895 | * @file: the file pointer | |
896 | * @cmd: either modify or retrieve lock state, possibly wait | |
897 | * @fl: type and range of lock | |
898 | * | |
899 | * Returns: errno | |
900 | */ | |
901 | ||
902 | static int gfs2_flock(struct file *file, int cmd, struct file_lock *fl) | |
903 | { | |
904 | struct gfs2_inode *ip = get_v2ip(file->f_mapping->host); | |
905 | struct gfs2_sbd *sdp = ip->i_sbd; | |
906 | ||
907 | atomic_inc(&ip->i_sbd->sd_ops_file); | |
908 | ||
909 | if (!(fl->fl_flags & FL_FLOCK)) | |
910 | return -ENOLCK; | |
911 | if ((ip->i_di.di_mode & (S_ISGID | S_IXGRP)) == S_ISGID) | |
912 | return -ENOLCK; | |
913 | ||
914 | if (sdp->sd_args.ar_localflocks) | |
915 | return flock_lock_file_wait(file, fl); | |
916 | ||
917 | if (fl->fl_type == F_UNLCK) { | |
918 | do_unflock(file, fl); | |
919 | return 0; | |
920 | } else | |
921 | return do_flock(file, cmd, fl); | |
922 | } | |
923 | ||
924 | struct file_operations gfs2_file_fops = { | |
925 | .llseek = gfs2_llseek, | |
926 | .read = gfs2_read, | |
18ec7d5c SW |
927 | .readv = gfs2_file_readv, |
928 | .aio_read = gfs2_file_aio_read, | |
929 | .write = generic_file_write, | |
930 | .writev = generic_file_writev, | |
931 | .aio_write = generic_file_aio_write, | |
b3b94faa DT |
932 | .ioctl = gfs2_ioctl, |
933 | .mmap = gfs2_mmap, | |
934 | .open = gfs2_open, | |
935 | .release = gfs2_close, | |
936 | .fsync = gfs2_fsync, | |
937 | .lock = gfs2_lock, | |
938 | .sendfile = gfs2_sendfile, | |
939 | .flock = gfs2_flock, | |
940 | }; | |
941 | ||
942 | struct file_operations gfs2_dir_fops = { | |
943 | .readdir = gfs2_readdir, | |
944 | .ioctl = gfs2_ioctl, | |
945 | .open = gfs2_open, | |
946 | .release = gfs2_close, | |
947 | .fsync = gfs2_fsync, | |
948 | .lock = gfs2_lock, | |
949 | .flock = gfs2_flock, | |
950 | }; | |
951 |