]>
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> | |
18ec7d5c | 20 | #include <linux/fs.h> |
5c676f6d | 21 | #include <linux/gfs2_ondisk.h> |
71b86f56 SW |
22 | #include <linux/ext2_fs.h> |
23 | #include <linux/crc32.h> | |
b3b94faa DT |
24 | #include <asm/semaphore.h> |
25 | #include <asm/uaccess.h> | |
26 | ||
27 | #include "gfs2.h" | |
5c676f6d SW |
28 | #include "lm_interface.h" |
29 | #include "incore.h" | |
b3b94faa DT |
30 | #include "bmap.h" |
31 | #include "dir.h" | |
32 | #include "glock.h" | |
33 | #include "glops.h" | |
34 | #include "inode.h" | |
b3b94faa DT |
35 | #include "lm.h" |
36 | #include "log.h" | |
37 | #include "meta_io.h" | |
38 | #include "ops_file.h" | |
39 | #include "ops_vm.h" | |
40 | #include "quota.h" | |
41 | #include "rgrp.h" | |
42 | #include "trans.h" | |
5c676f6d | 43 | #include "util.h" |
71b86f56 | 44 | #include "eaops.h" |
b3b94faa DT |
45 | |
46 | /* "bad" is for NFS support */ | |
47 | struct filldir_bad_entry { | |
48 | char *fbe_name; | |
49 | unsigned int fbe_length; | |
50 | uint64_t fbe_offset; | |
51 | struct gfs2_inum fbe_inum; | |
52 | unsigned int fbe_type; | |
53 | }; | |
54 | ||
55 | struct filldir_bad { | |
56 | struct gfs2_sbd *fdb_sbd; | |
57 | ||
58 | struct filldir_bad_entry *fdb_entry; | |
59 | unsigned int fdb_entry_num; | |
60 | unsigned int fdb_entry_off; | |
61 | ||
62 | char *fdb_name; | |
63 | unsigned int fdb_name_size; | |
64 | unsigned int fdb_name_off; | |
65 | }; | |
66 | ||
67 | /* For regular, non-NFS */ | |
68 | struct filldir_reg { | |
69 | struct gfs2_sbd *fdr_sbd; | |
70 | int fdr_prefetch; | |
71 | ||
72 | filldir_t fdr_filldir; | |
73 | void *fdr_opaque; | |
74 | }; | |
75 | ||
61a30dcb SW |
76 | /* |
77 | * Most fields left uninitialised to catch anybody who tries to | |
78 | * use them. f_flags set to prevent file_accessed() from touching | |
79 | * any other part of this. Its use is purely as a flag so that we | |
80 | * know (in readpage()) whether or not do to locking. | |
81 | */ | |
82 | struct file gfs2_internal_file_sentinal = { | |
83 | .f_flags = O_NOATIME|O_RDONLY, | |
84 | }; | |
85 | ||
18ec7d5c SW |
86 | static int gfs2_read_actor(read_descriptor_t *desc, struct page *page, |
87 | unsigned long offset, unsigned long size) | |
88 | { | |
89 | char *kaddr; | |
90 | unsigned long count = desc->count; | |
91 | ||
92 | if (size > count) | |
93 | size = count; | |
94 | ||
95 | kaddr = kmap(page); | |
96 | memcpy(desc->arg.buf, kaddr + offset, size); | |
97 | kunmap(page); | |
98 | ||
99 | desc->count = count - size; | |
100 | desc->written += size; | |
101 | desc->arg.buf += size; | |
102 | return size; | |
103 | } | |
104 | ||
105 | int gfs2_internal_read(struct gfs2_inode *ip, struct file_ra_state *ra_state, | |
106 | char *buf, loff_t *pos, unsigned size) | |
107 | { | |
108 | struct inode *inode = ip->i_vnode; | |
109 | read_descriptor_t desc; | |
110 | desc.written = 0; | |
111 | desc.arg.buf = buf; | |
112 | desc.count = size; | |
113 | desc.error = 0; | |
61a30dcb SW |
114 | do_generic_mapping_read(inode->i_mapping, ra_state, |
115 | &gfs2_internal_file_sentinal, pos, &desc, | |
116 | gfs2_read_actor); | |
18ec7d5c SW |
117 | return desc.written ? desc.written : desc.error; |
118 | } | |
b3b94faa DT |
119 | |
120 | /** | |
121 | * gfs2_llseek - seek to a location in a file | |
122 | * @file: the file | |
123 | * @offset: the offset | |
124 | * @origin: Where to seek from (SEEK_SET, SEEK_CUR, or SEEK_END) | |
125 | * | |
126 | * SEEK_END requires the glock for the file because it references the | |
127 | * file's size. | |
128 | * | |
129 | * Returns: The new offset, or errno | |
130 | */ | |
131 | ||
132 | static loff_t gfs2_llseek(struct file *file, loff_t offset, int origin) | |
133 | { | |
5c676f6d | 134 | struct gfs2_inode *ip = file->f_mapping->host->u.generic_ip; |
b3b94faa DT |
135 | struct gfs2_holder i_gh; |
136 | loff_t error; | |
137 | ||
b3b94faa DT |
138 | if (origin == 2) { |
139 | error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED, LM_FLAG_ANY, | |
140 | &i_gh); | |
141 | if (!error) { | |
142 | error = remote_llseek(file, offset, origin); | |
143 | gfs2_glock_dq_uninit(&i_gh); | |
144 | } | |
145 | } else | |
146 | error = remote_llseek(file, offset, origin); | |
147 | ||
148 | return error; | |
149 | } | |
150 | ||
b3b94faa | 151 | |
18ec7d5c SW |
152 | static ssize_t gfs2_direct_IO_read(struct kiocb *iocb, const struct iovec *iov, |
153 | loff_t offset, unsigned long nr_segs) | |
b3b94faa | 154 | { |
18ec7d5c SW |
155 | struct file *file = iocb->ki_filp; |
156 | struct address_space *mapping = file->f_mapping; | |
157 | ssize_t retval; | |
b3b94faa | 158 | |
18ec7d5c SW |
159 | retval = filemap_write_and_wait(mapping); |
160 | if (retval == 0) { | |
161 | retval = mapping->a_ops->direct_IO(READ, iocb, iov, offset, | |
162 | nr_segs); | |
b3b94faa | 163 | } |
18ec7d5c | 164 | return retval; |
b3b94faa DT |
165 | } |
166 | ||
167 | /** | |
18ec7d5c SW |
168 | * __gfs2_file_aio_read - The main GFS2 read function |
169 | * | |
170 | * N.B. This is almost, but not quite the same as __generic_file_aio_read() | |
171 | * the important subtle different being that inode->i_size isn't valid | |
172 | * unless we are holding a lock, and we do this _only_ on the O_DIRECT | |
173 | * path since otherwise locking is done entirely at the page cache | |
174 | * layer. | |
b3b94faa | 175 | */ |
18ec7d5c SW |
176 | static ssize_t __gfs2_file_aio_read(struct kiocb *iocb, |
177 | const struct iovec *iov, | |
178 | unsigned long nr_segs, loff_t *ppos) | |
b3b94faa | 179 | { |
18ec7d5c | 180 | struct file *filp = iocb->ki_filp; |
5c676f6d | 181 | struct gfs2_inode *ip = filp->f_mapping->host->u.generic_ip; |
b3b94faa | 182 | struct gfs2_holder gh; |
18ec7d5c SW |
183 | ssize_t retval; |
184 | unsigned long seg; | |
185 | size_t count; | |
186 | ||
187 | count = 0; | |
188 | for (seg = 0; seg < nr_segs; seg++) { | |
189 | const struct iovec *iv = &iov[seg]; | |
190 | ||
191 | /* | |
192 | * If any segment has a negative length, or the cumulative | |
193 | * length ever wraps negative then return -EINVAL. | |
194 | */ | |
d1665e41 SW |
195 | count += iv->iov_len; |
196 | if (unlikely((ssize_t)(count|iv->iov_len) < 0)) | |
197 | return -EINVAL; | |
198 | if (access_ok(VERIFY_WRITE, iv->iov_base, iv->iov_len)) | |
199 | continue; | |
200 | if (seg == 0) | |
201 | return -EFAULT; | |
202 | nr_segs = seg; | |
203 | count -= iv->iov_len; /* This segment is no good */ | |
204 | break; | |
18ec7d5c SW |
205 | } |
206 | ||
207 | /* coalesce the iovecs and go direct-to-BIO for O_DIRECT */ | |
208 | if (filp->f_flags & O_DIRECT) { | |
209 | loff_t pos = *ppos, size; | |
210 | struct address_space *mapping; | |
211 | struct inode *inode; | |
212 | ||
213 | mapping = filp->f_mapping; | |
214 | inode = mapping->host; | |
215 | retval = 0; | |
216 | if (!count) | |
217 | goto out; /* skip atime */ | |
218 | ||
219 | gfs2_holder_init(ip->i_gl, LM_ST_SHARED, GL_ATIME, &gh); | |
220 | retval = gfs2_glock_nq_m_atime(1, &gh); | |
221 | if (retval) | |
222 | goto out; | |
d1665e41 SW |
223 | if (gfs2_is_stuffed(ip)) { |
224 | gfs2_glock_dq_m(1, &gh); | |
225 | gfs2_holder_uninit(&gh); | |
226 | goto fallback_to_normal; | |
227 | } | |
18ec7d5c SW |
228 | size = i_size_read(inode); |
229 | if (pos < size) { | |
d1665e41 | 230 | retval = gfs2_direct_IO_read(iocb, iov, pos, nr_segs); |
18ec7d5c SW |
231 | if (retval > 0 && !is_sync_kiocb(iocb)) |
232 | retval = -EIOCBQUEUED; | |
233 | if (retval > 0) | |
234 | *ppos = pos + retval; | |
b3b94faa | 235 | } |
18ec7d5c SW |
236 | file_accessed(filp); |
237 | gfs2_glock_dq_m(1, &gh); | |
238 | gfs2_holder_uninit(&gh); | |
b3b94faa | 239 | goto out; |
18ec7d5c | 240 | } |
b3b94faa | 241 | |
d1665e41 | 242 | fallback_to_normal: |
18ec7d5c SW |
243 | retval = 0; |
244 | if (count) { | |
245 | for (seg = 0; seg < nr_segs; seg++) { | |
246 | read_descriptor_t desc; | |
247 | ||
248 | desc.written = 0; | |
249 | desc.arg.buf = iov[seg].iov_base; | |
250 | desc.count = iov[seg].iov_len; | |
251 | if (desc.count == 0) | |
252 | continue; | |
253 | desc.error = 0; | |
254 | do_generic_file_read(filp,ppos,&desc,file_read_actor); | |
255 | retval += desc.written; | |
256 | if (desc.error) { | |
257 | retval = retval ?: desc.error; | |
258 | break; | |
259 | } | |
260 | } | |
261 | } | |
262 | out: | |
263 | return retval; | |
b3b94faa DT |
264 | } |
265 | ||
266 | /** | |
267 | * gfs2_read - Read bytes from a file | |
268 | * @file: The file to read from | |
269 | * @buf: The buffer to copy into | |
270 | * @size: The amount of data requested | |
271 | * @offset: The current file offset | |
272 | * | |
273 | * Outputs: Offset - updated according to number of bytes read | |
274 | * | |
275 | * Returns: The number of bytes read, errno on failure | |
276 | */ | |
277 | ||
18ec7d5c | 278 | static ssize_t gfs2_read(struct file *filp, char __user *buf, size_t size, |
b3b94faa DT |
279 | loff_t *offset) |
280 | { | |
b3b94faa | 281 | struct iovec local_iov = { .iov_base = buf, .iov_len = size }; |
18ec7d5c SW |
282 | struct kiocb kiocb; |
283 | ssize_t ret; | |
b3b94faa | 284 | |
18ec7d5c SW |
285 | init_sync_kiocb(&kiocb, filp); |
286 | ret = __gfs2_file_aio_read(&kiocb, &local_iov, 1, offset); | |
287 | if (-EIOCBQUEUED == ret) | |
288 | ret = wait_on_sync_kiocb(&kiocb); | |
289 | return ret; | |
b3b94faa DT |
290 | } |
291 | ||
18ec7d5c SW |
292 | static ssize_t gfs2_file_readv(struct file *filp, const struct iovec *iov, |
293 | unsigned long nr_segs, loff_t *ppos) | |
b3b94faa | 294 | { |
18ec7d5c SW |
295 | struct kiocb kiocb; |
296 | ssize_t ret; | |
b3b94faa | 297 | |
18ec7d5c SW |
298 | init_sync_kiocb(&kiocb, filp); |
299 | ret = __gfs2_file_aio_read(&kiocb, iov, nr_segs, ppos); | |
300 | if (-EIOCBQUEUED == ret) | |
301 | ret = wait_on_sync_kiocb(&kiocb); | |
302 | return ret; | |
b3b94faa DT |
303 | } |
304 | ||
18ec7d5c SW |
305 | static ssize_t gfs2_file_aio_read(struct kiocb *iocb, char __user *buf, |
306 | size_t count, loff_t pos) | |
b3b94faa | 307 | { |
18ec7d5c | 308 | struct iovec local_iov = { .iov_base = buf, .iov_len = count }; |
b3b94faa | 309 | |
18ec7d5c SW |
310 | BUG_ON(iocb->ki_pos != pos); |
311 | return __gfs2_file_aio_read(iocb, &local_iov, 1, &iocb->ki_pos); | |
b3b94faa DT |
312 | } |
313 | ||
b3b94faa DT |
314 | |
315 | /** | |
316 | * filldir_reg_func - Report a directory entry to the caller of gfs2_dir_read() | |
317 | * @opaque: opaque data used by the function | |
318 | * @name: the name of the directory entry | |
319 | * @length: the length of the name | |
320 | * @offset: the entry's offset in the directory | |
321 | * @inum: the inode number the entry points to | |
322 | * @type: the type of inode the entry points to | |
323 | * | |
324 | * Returns: 0 on success, 1 if buffer full | |
325 | */ | |
326 | ||
327 | static int filldir_reg_func(void *opaque, const char *name, unsigned int length, | |
328 | uint64_t offset, struct gfs2_inum *inum, | |
329 | unsigned int type) | |
330 | { | |
331 | struct filldir_reg *fdr = (struct filldir_reg *)opaque; | |
332 | struct gfs2_sbd *sdp = fdr->fdr_sbd; | |
333 | int error; | |
334 | ||
335 | error = fdr->fdr_filldir(fdr->fdr_opaque, name, length, offset, | |
336 | inum->no_formal_ino, type); | |
337 | if (error) | |
338 | return 1; | |
339 | ||
340 | if (fdr->fdr_prefetch && !(length == 1 && *name == '.')) { | |
341 | gfs2_glock_prefetch_num(sdp, | |
342 | inum->no_addr, &gfs2_inode_glops, | |
343 | LM_ST_SHARED, LM_FLAG_TRY | LM_FLAG_ANY); | |
344 | gfs2_glock_prefetch_num(sdp, | |
345 | inum->no_addr, &gfs2_iopen_glops, | |
346 | LM_ST_SHARED, LM_FLAG_TRY); | |
347 | } | |
348 | ||
349 | return 0; | |
350 | } | |
351 | ||
352 | /** | |
353 | * readdir_reg - Read directory entries from a directory | |
354 | * @file: The directory to read from | |
355 | * @dirent: Buffer for dirents | |
356 | * @filldir: Function used to do the copying | |
357 | * | |
358 | * Returns: errno | |
359 | */ | |
360 | ||
361 | static int readdir_reg(struct file *file, void *dirent, filldir_t filldir) | |
362 | { | |
71b86f56 SW |
363 | struct inode *dir = file->f_mapping->host; |
364 | struct gfs2_inode *dip = dir->u.generic_ip; | |
b3b94faa DT |
365 | struct filldir_reg fdr; |
366 | struct gfs2_holder d_gh; | |
367 | uint64_t offset = file->f_pos; | |
368 | int error; | |
369 | ||
370 | fdr.fdr_sbd = dip->i_sbd; | |
371 | fdr.fdr_prefetch = 1; | |
372 | fdr.fdr_filldir = filldir; | |
373 | fdr.fdr_opaque = dirent; | |
374 | ||
375 | gfs2_holder_init(dip->i_gl, LM_ST_SHARED, GL_ATIME, &d_gh); | |
376 | error = gfs2_glock_nq_atime(&d_gh); | |
377 | if (error) { | |
378 | gfs2_holder_uninit(&d_gh); | |
379 | return error; | |
380 | } | |
381 | ||
71b86f56 | 382 | error = gfs2_dir_read(dir, &offset, &fdr, filldir_reg_func); |
b3b94faa DT |
383 | |
384 | gfs2_glock_dq_uninit(&d_gh); | |
385 | ||
386 | file->f_pos = offset; | |
387 | ||
388 | return error; | |
389 | } | |
390 | ||
391 | /** | |
392 | * filldir_bad_func - Report a directory entry to the caller of gfs2_dir_read() | |
393 | * @opaque: opaque data used by the function | |
394 | * @name: the name of the directory entry | |
395 | * @length: the length of the name | |
396 | * @offset: the entry's offset in the directory | |
397 | * @inum: the inode number the entry points to | |
398 | * @type: the type of inode the entry points to | |
399 | * | |
400 | * For supporting NFS. | |
401 | * | |
402 | * Returns: 0 on success, 1 if buffer full | |
403 | */ | |
404 | ||
405 | static int filldir_bad_func(void *opaque, const char *name, unsigned int length, | |
406 | uint64_t offset, struct gfs2_inum *inum, | |
407 | unsigned int type) | |
408 | { | |
409 | struct filldir_bad *fdb = (struct filldir_bad *)opaque; | |
410 | struct gfs2_sbd *sdp = fdb->fdb_sbd; | |
411 | struct filldir_bad_entry *fbe; | |
412 | ||
413 | if (fdb->fdb_entry_off == fdb->fdb_entry_num || | |
414 | fdb->fdb_name_off + length > fdb->fdb_name_size) | |
415 | return 1; | |
416 | ||
417 | fbe = &fdb->fdb_entry[fdb->fdb_entry_off]; | |
418 | fbe->fbe_name = fdb->fdb_name + fdb->fdb_name_off; | |
419 | memcpy(fbe->fbe_name, name, length); | |
420 | fbe->fbe_length = length; | |
421 | fbe->fbe_offset = offset; | |
422 | fbe->fbe_inum = *inum; | |
423 | fbe->fbe_type = type; | |
424 | ||
425 | fdb->fdb_entry_off++; | |
426 | fdb->fdb_name_off += length; | |
427 | ||
428 | if (!(length == 1 && *name == '.')) { | |
429 | gfs2_glock_prefetch_num(sdp, | |
430 | inum->no_addr, &gfs2_inode_glops, | |
431 | LM_ST_SHARED, LM_FLAG_TRY | LM_FLAG_ANY); | |
432 | gfs2_glock_prefetch_num(sdp, | |
433 | inum->no_addr, &gfs2_iopen_glops, | |
434 | LM_ST_SHARED, LM_FLAG_TRY); | |
435 | } | |
436 | ||
437 | return 0; | |
438 | } | |
439 | ||
440 | /** | |
441 | * readdir_bad - Read directory entries from a directory | |
442 | * @file: The directory to read from | |
443 | * @dirent: Buffer for dirents | |
444 | * @filldir: Function used to do the copying | |
445 | * | |
446 | * For supporting NFS. | |
447 | * | |
448 | * Returns: errno | |
449 | */ | |
450 | ||
451 | static int readdir_bad(struct file *file, void *dirent, filldir_t filldir) | |
452 | { | |
71b86f56 SW |
453 | struct inode *dir = file->f_mapping->host; |
454 | struct gfs2_inode *dip = dir->u.generic_ip; | |
b3b94faa DT |
455 | struct gfs2_sbd *sdp = dip->i_sbd; |
456 | struct filldir_reg fdr; | |
457 | unsigned int entries, size; | |
458 | struct filldir_bad *fdb; | |
459 | struct gfs2_holder d_gh; | |
460 | uint64_t offset = file->f_pos; | |
461 | unsigned int x; | |
462 | struct filldir_bad_entry *fbe; | |
463 | int error; | |
464 | ||
465 | entries = gfs2_tune_get(sdp, gt_entries_per_readdir); | |
466 | size = sizeof(struct filldir_bad) + | |
467 | entries * (sizeof(struct filldir_bad_entry) + GFS2_FAST_NAME_SIZE); | |
468 | ||
469 | fdb = kzalloc(size, GFP_KERNEL); | |
470 | if (!fdb) | |
471 | return -ENOMEM; | |
472 | ||
473 | fdb->fdb_sbd = sdp; | |
474 | fdb->fdb_entry = (struct filldir_bad_entry *)(fdb + 1); | |
475 | fdb->fdb_entry_num = entries; | |
476 | fdb->fdb_name = ((char *)fdb) + sizeof(struct filldir_bad) + | |
477 | entries * sizeof(struct filldir_bad_entry); | |
478 | fdb->fdb_name_size = entries * GFS2_FAST_NAME_SIZE; | |
479 | ||
480 | gfs2_holder_init(dip->i_gl, LM_ST_SHARED, GL_ATIME, &d_gh); | |
481 | error = gfs2_glock_nq_atime(&d_gh); | |
482 | if (error) { | |
483 | gfs2_holder_uninit(&d_gh); | |
484 | goto out; | |
485 | } | |
486 | ||
71b86f56 | 487 | error = gfs2_dir_read(dir, &offset, fdb, filldir_bad_func); |
b3b94faa DT |
488 | |
489 | gfs2_glock_dq_uninit(&d_gh); | |
490 | ||
491 | fdr.fdr_sbd = sdp; | |
492 | fdr.fdr_prefetch = 0; | |
493 | fdr.fdr_filldir = filldir; | |
494 | fdr.fdr_opaque = dirent; | |
495 | ||
496 | for (x = 0; x < fdb->fdb_entry_off; x++) { | |
497 | fbe = &fdb->fdb_entry[x]; | |
498 | ||
499 | error = filldir_reg_func(&fdr, | |
500 | fbe->fbe_name, fbe->fbe_length, | |
501 | fbe->fbe_offset, | |
502 | &fbe->fbe_inum, fbe->fbe_type); | |
503 | if (error) { | |
504 | file->f_pos = fbe->fbe_offset; | |
505 | error = 0; | |
506 | goto out; | |
507 | } | |
508 | } | |
509 | ||
510 | file->f_pos = offset; | |
511 | ||
512 | out: | |
513 | kfree(fdb); | |
514 | ||
515 | return error; | |
516 | } | |
517 | ||
518 | /** | |
519 | * gfs2_readdir - Read directory entries from a directory | |
520 | * @file: The directory to read from | |
521 | * @dirent: Buffer for dirents | |
522 | * @filldir: Function used to do the copying | |
523 | * | |
524 | * Returns: errno | |
525 | */ | |
526 | ||
527 | static int gfs2_readdir(struct file *file, void *dirent, filldir_t filldir) | |
528 | { | |
529 | int error; | |
530 | ||
b3b94faa DT |
531 | if (strcmp(current->comm, "nfsd") != 0) |
532 | error = readdir_reg(file, dirent, filldir); | |
533 | else | |
534 | error = readdir_bad(file, dirent, filldir); | |
535 | ||
536 | return error; | |
537 | } | |
538 | ||
71b86f56 SW |
539 | const struct gfs2_flag_eattr { |
540 | u32 flag; | |
541 | u32 ext2; | |
542 | } gfs2_flag_eattrs[] = { | |
543 | { | |
544 | .flag = GFS2_DIF_IMMUTABLE, | |
545 | .ext2 = EXT2_IMMUTABLE_FL, | |
546 | }, { | |
547 | .flag = GFS2_DIF_APPENDONLY, | |
548 | .ext2 = EXT2_APPEND_FL, | |
549 | }, { | |
550 | .flag = GFS2_DIF_JDATA, | |
551 | .ext2 = EXT2_JOURNAL_DATA_FL, | |
552 | }, { | |
553 | .flag = GFS2_DIF_EXHASH, | |
554 | .ext2 = EXT2_INDEX_FL, | |
555 | }, { | |
556 | .flag = GFS2_DIF_EA_INDIRECT, | |
557 | }, { | |
558 | .flag = GFS2_DIF_DIRECTIO, | |
559 | }, { | |
560 | .flag = GFS2_DIF_NOATIME, | |
561 | .ext2 = EXT2_NOATIME_FL, | |
562 | }, { | |
563 | .flag = GFS2_DIF_SYNC, | |
564 | .ext2 = EXT2_SYNC_FL, | |
565 | }, { | |
566 | .flag = GFS2_DIF_SYSTEM, | |
567 | }, { | |
568 | .flag = GFS2_DIF_TRUNC_IN_PROG, | |
569 | }, { | |
570 | .flag = GFS2_DIF_INHERIT_JDATA, | |
571 | }, { | |
572 | .flag = GFS2_DIF_INHERIT_DIRECTIO, | |
573 | }, { | |
574 | }, | |
575 | }; | |
576 | ||
577 | static const struct gfs2_flag_eattr *get_by_ext2(u32 ext2) | |
578 | { | |
579 | const struct gfs2_flag_eattr *p = gfs2_flag_eattrs; | |
580 | for(; p->flag; p++) { | |
581 | if (ext2 == p->ext2) | |
582 | return p; | |
583 | } | |
584 | return NULL; | |
585 | } | |
586 | ||
587 | static const struct gfs2_flag_eattr *get_by_gfs2(u32 gfs2) | |
588 | { | |
589 | const struct gfs2_flag_eattr *p = gfs2_flag_eattrs; | |
590 | for(; p->flag; p++) { | |
591 | if (gfs2 == p->flag) | |
592 | return p; | |
593 | } | |
594 | return NULL; | |
595 | } | |
596 | ||
597 | static u32 gfs2_flags_to_ext2(u32 gfs2) | |
598 | { | |
599 | const struct gfs2_flag_eattr *ea; | |
600 | u32 ext2 = 0; | |
601 | u32 mask = 1; | |
602 | ||
603 | for(; mask != 0; mask <<=1) { | |
604 | if (mask & gfs2) { | |
605 | ea = get_by_gfs2(mask); | |
606 | if (ea) | |
607 | ext2 |= ea->ext2; | |
608 | } | |
609 | } | |
610 | return ext2; | |
611 | } | |
612 | ||
613 | static int gfs2_flags_from_ext2(u32 *gfs2, u32 ext2) | |
614 | { | |
615 | const struct gfs2_flag_eattr *ea; | |
616 | u32 mask = 1; | |
617 | ||
618 | for(; mask != 0; mask <<= 1) { | |
619 | if (mask & ext2) { | |
620 | ea = get_by_ext2(mask); | |
621 | if (ea == NULL) | |
622 | return -EINVAL; | |
623 | *gfs2 |= ea->flag; | |
624 | } | |
625 | } | |
626 | return 0; | |
627 | } | |
628 | ||
629 | static int get_ext2_flags(struct inode *inode, u32 __user *ptr) | |
630 | { | |
631 | struct gfs2_inode *ip = inode->u.generic_ip; | |
632 | struct gfs2_holder gh; | |
633 | int error; | |
634 | u32 ext2; | |
635 | ||
636 | gfs2_holder_init(ip->i_gl, LM_ST_SHARED, GL_ATIME, &gh); | |
637 | error = gfs2_glock_nq_m_atime(1, &gh); | |
638 | if (error) | |
639 | return error; | |
640 | ||
641 | ext2 = gfs2_flags_to_ext2(ip->i_di.di_flags); | |
642 | if (put_user(ext2, ptr)) | |
643 | error = -EFAULT; | |
644 | ||
645 | gfs2_glock_dq_m(1, &gh); | |
646 | gfs2_holder_uninit(&gh); | |
647 | return error; | |
648 | } | |
649 | ||
650 | /* Flags that can be set by user space */ | |
651 | #define GFS2_FLAGS_USER_SET (GFS2_DIF_JDATA| \ | |
652 | GFS2_DIF_DIRECTIO| \ | |
653 | GFS2_DIF_IMMUTABLE| \ | |
654 | GFS2_DIF_APPENDONLY| \ | |
655 | GFS2_DIF_NOATIME| \ | |
656 | GFS2_DIF_SYNC| \ | |
657 | GFS2_DIF_SYSTEM| \ | |
658 | GFS2_DIF_INHERIT_DIRECTIO| \ | |
659 | GFS2_DIF_INHERIT_JDATA) | |
660 | ||
661 | /** | |
662 | * gfs2_set_flags - set flags on an inode | |
663 | * @inode: The inode | |
664 | * @flags: The flags to set | |
665 | * @mask: Indicates which flags are valid | |
666 | * | |
667 | */ | |
668 | static int gfs2_set_flags(struct inode *inode, u32 flags, u32 mask) | |
669 | { | |
670 | struct gfs2_inode *ip = inode->u.generic_ip; | |
671 | struct buffer_head *bh; | |
672 | struct gfs2_holder gh; | |
673 | int error; | |
674 | u32 new_flags; | |
675 | ||
676 | gfs2_holder_init(ip->i_gl, LM_ST_EXCLUSIVE, 0, &gh); | |
677 | error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0, &gh); | |
678 | if (error) | |
679 | return error; | |
680 | ||
681 | new_flags = (ip->i_di.di_flags & ~mask) | (flags & mask); | |
682 | if ((new_flags ^ flags) == 0) | |
683 | goto out; | |
684 | ||
685 | error = -EINVAL; | |
686 | if ((new_flags ^ flags) & ~GFS2_FLAGS_USER_SET) | |
687 | goto out; | |
688 | ||
689 | if (S_ISDIR(inode->i_mode)) { | |
690 | if ((new_flags ^ flags) & (GFS2_DIF_JDATA | GFS2_DIF_DIRECTIO)) | |
691 | goto out; | |
692 | } else if (S_ISREG(inode->i_mode)) { | |
693 | if ((new_flags ^ flags) & (GFS2_DIF_INHERIT_DIRECTIO| | |
694 | GFS2_DIF_INHERIT_JDATA)) | |
695 | goto out; | |
696 | } else | |
697 | goto out; | |
698 | ||
699 | error = -EPERM; | |
700 | if (IS_IMMUTABLE(inode) && (new_flags & GFS2_DIF_IMMUTABLE)) | |
701 | goto out; | |
702 | if (IS_APPEND(inode) && (new_flags & GFS2_DIF_APPENDONLY)) | |
703 | goto out; | |
704 | error = gfs2_repermission(inode, MAY_WRITE, NULL); | |
705 | if (error) | |
706 | goto out; | |
707 | ||
708 | error = gfs2_meta_inode_buffer(ip, &bh); | |
709 | if (error) | |
710 | goto out; | |
711 | gfs2_trans_add_bh(ip->i_gl, bh, 1); | |
712 | ip->i_di.di_flags = new_flags; | |
713 | gfs2_dinode_out(&ip->i_di, bh->b_data); | |
714 | brelse(bh); | |
715 | out: | |
716 | gfs2_glock_dq_uninit(&gh); | |
717 | return error; | |
718 | } | |
719 | ||
720 | static int set_ext2_flags(struct inode *inode, u32 __user *ptr) | |
721 | { | |
722 | u32 ext2, gfs2; | |
723 | if (get_user(ext2, ptr)) | |
724 | return -EFAULT; | |
725 | if (gfs2_flags_from_ext2(&gfs2, ext2)) | |
726 | return -EINVAL; | |
727 | return gfs2_set_flags(inode, gfs2, ~0); | |
728 | } | |
729 | ||
730 | int gfs2_ioctl(struct inode *inode, struct file *filp, unsigned int cmd, | |
731 | unsigned long arg) | |
732 | { | |
733 | switch(cmd) { | |
734 | case EXT2_IOC_GETFLAGS: | |
735 | return get_ext2_flags(inode, (u32 __user *)arg); | |
736 | case EXT2_IOC_SETFLAGS: | |
737 | return set_ext2_flags(inode, (u32 __user *)arg); | |
738 | } | |
739 | return -ENOTTY; | |
740 | } | |
741 | ||
742 | ||
b3b94faa DT |
743 | /** |
744 | * gfs2_mmap - | |
745 | * @file: The file to map | |
746 | * @vma: The VMA which described the mapping | |
747 | * | |
748 | * Returns: 0 or error code | |
749 | */ | |
750 | ||
751 | static int gfs2_mmap(struct file *file, struct vm_area_struct *vma) | |
752 | { | |
5c676f6d | 753 | struct gfs2_inode *ip = file->f_mapping->host->u.generic_ip; |
b3b94faa DT |
754 | struct gfs2_holder i_gh; |
755 | int error; | |
756 | ||
b3b94faa DT |
757 | gfs2_holder_init(ip->i_gl, LM_ST_SHARED, GL_ATIME, &i_gh); |
758 | error = gfs2_glock_nq_atime(&i_gh); | |
759 | if (error) { | |
760 | gfs2_holder_uninit(&i_gh); | |
761 | return error; | |
762 | } | |
763 | ||
18ec7d5c SW |
764 | /* This is VM_MAYWRITE instead of VM_WRITE because a call |
765 | to mprotect() can turn on VM_WRITE later. */ | |
766 | ||
767 | if ((vma->vm_flags & (VM_MAYSHARE | VM_MAYWRITE)) == | |
768 | (VM_MAYSHARE | VM_MAYWRITE)) | |
769 | vma->vm_ops = &gfs2_vm_ops_sharewrite; | |
770 | else | |
771 | vma->vm_ops = &gfs2_vm_ops_private; | |
b3b94faa DT |
772 | |
773 | gfs2_glock_dq_uninit(&i_gh); | |
774 | ||
775 | return error; | |
776 | } | |
777 | ||
778 | /** | |
779 | * gfs2_open - open a file | |
780 | * @inode: the inode to open | |
781 | * @file: the struct file for this opening | |
782 | * | |
783 | * Returns: errno | |
784 | */ | |
785 | ||
786 | static int gfs2_open(struct inode *inode, struct file *file) | |
787 | { | |
5c676f6d | 788 | struct gfs2_inode *ip = inode->u.generic_ip; |
b3b94faa DT |
789 | struct gfs2_holder i_gh; |
790 | struct gfs2_file *fp; | |
791 | int error; | |
792 | ||
b3b94faa DT |
793 | fp = kzalloc(sizeof(struct gfs2_file), GFP_KERNEL); |
794 | if (!fp) | |
795 | return -ENOMEM; | |
796 | ||
f55ab26a | 797 | mutex_init(&fp->f_fl_mutex); |
b3b94faa DT |
798 | |
799 | fp->f_inode = ip; | |
800 | fp->f_vfile = file; | |
801 | ||
5c676f6d SW |
802 | gfs2_assert_warn(ip->i_sbd, !file->private_data); |
803 | file->private_data = fp; | |
b3b94faa DT |
804 | |
805 | if (S_ISREG(ip->i_di.di_mode)) { | |
806 | error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED, LM_FLAG_ANY, | |
807 | &i_gh); | |
808 | if (error) | |
809 | goto fail; | |
810 | ||
811 | if (!(file->f_flags & O_LARGEFILE) && | |
812 | ip->i_di.di_size > MAX_NON_LFS) { | |
813 | error = -EFBIG; | |
814 | goto fail_gunlock; | |
815 | } | |
816 | ||
817 | /* Listen to the Direct I/O flag */ | |
818 | ||
819 | if (ip->i_di.di_flags & GFS2_DIF_DIRECTIO) | |
820 | file->f_flags |= O_DIRECT; | |
821 | ||
b3b94faa DT |
822 | gfs2_glock_dq_uninit(&i_gh); |
823 | } | |
824 | ||
825 | return 0; | |
826 | ||
827 | fail_gunlock: | |
828 | gfs2_glock_dq_uninit(&i_gh); | |
829 | ||
830 | fail: | |
5c676f6d | 831 | file->private_data = NULL; |
b3b94faa DT |
832 | kfree(fp); |
833 | ||
834 | return error; | |
835 | } | |
836 | ||
837 | /** | |
838 | * gfs2_close - called to close a struct file | |
839 | * @inode: the inode the struct file belongs to | |
840 | * @file: the struct file being closed | |
841 | * | |
842 | * Returns: errno | |
843 | */ | |
844 | ||
845 | static int gfs2_close(struct inode *inode, struct file *file) | |
846 | { | |
5c676f6d | 847 | struct gfs2_sbd *sdp = inode->i_sb->s_fs_info; |
b3b94faa DT |
848 | struct gfs2_file *fp; |
849 | ||
5c676f6d SW |
850 | fp = file->private_data; |
851 | file->private_data = NULL; | |
b3b94faa DT |
852 | |
853 | if (gfs2_assert_warn(sdp, fp)) | |
854 | return -EIO; | |
855 | ||
856 | kfree(fp); | |
857 | ||
858 | return 0; | |
859 | } | |
860 | ||
861 | /** | |
862 | * gfs2_fsync - sync the dirty data for a file (across the cluster) | |
863 | * @file: the file that points to the dentry (we ignore this) | |
864 | * @dentry: the dentry that points to the inode to sync | |
865 | * | |
866 | * Returns: errno | |
867 | */ | |
868 | ||
869 | static int gfs2_fsync(struct file *file, struct dentry *dentry, int datasync) | |
870 | { | |
5c676f6d | 871 | struct gfs2_inode *ip = dentry->d_inode->u.generic_ip; |
b3b94faa | 872 | |
b3b94faa DT |
873 | gfs2_log_flush_glock(ip->i_gl); |
874 | ||
875 | return 0; | |
876 | } | |
877 | ||
878 | /** | |
879 | * gfs2_lock - acquire/release a posix lock on a file | |
880 | * @file: the file pointer | |
881 | * @cmd: either modify or retrieve lock state, possibly wait | |
882 | * @fl: type and range of lock | |
883 | * | |
884 | * Returns: errno | |
885 | */ | |
886 | ||
887 | static int gfs2_lock(struct file *file, int cmd, struct file_lock *fl) | |
888 | { | |
5c676f6d | 889 | struct gfs2_inode *ip = file->f_mapping->host->u.generic_ip; |
b3b94faa DT |
890 | struct gfs2_sbd *sdp = ip->i_sbd; |
891 | struct lm_lockname name = | |
892 | { .ln_number = ip->i_num.no_addr, | |
893 | .ln_type = LM_TYPE_PLOCK }; | |
894 | ||
b3b94faa DT |
895 | if (!(fl->fl_flags & FL_POSIX)) |
896 | return -ENOLCK; | |
897 | if ((ip->i_di.di_mode & (S_ISGID | S_IXGRP)) == S_ISGID) | |
898 | return -ENOLCK; | |
899 | ||
900 | if (sdp->sd_args.ar_localflocks) { | |
901 | if (IS_GETLK(cmd)) { | |
902 | struct file_lock *tmp; | |
903 | lock_kernel(); | |
904 | tmp = posix_test_lock(file, fl); | |
905 | fl->fl_type = F_UNLCK; | |
906 | if (tmp) | |
907 | memcpy(fl, tmp, sizeof(struct file_lock)); | |
908 | unlock_kernel(); | |
909 | return 0; | |
910 | } else { | |
911 | int error; | |
912 | lock_kernel(); | |
913 | error = posix_lock_file_wait(file, fl); | |
914 | unlock_kernel(); | |
915 | return error; | |
916 | } | |
917 | } | |
918 | ||
919 | if (IS_GETLK(cmd)) | |
920 | return gfs2_lm_plock_get(sdp, &name, file, fl); | |
921 | else if (fl->fl_type == F_UNLCK) | |
922 | return gfs2_lm_punlock(sdp, &name, file, fl); | |
923 | else | |
924 | return gfs2_lm_plock(sdp, &name, file, cmd, fl); | |
925 | } | |
926 | ||
927 | /** | |
928 | * gfs2_sendfile - Send bytes to a file or socket | |
929 | * @in_file: The file to read from | |
930 | * @out_file: The file to write to | |
931 | * @count: The amount of data | |
932 | * @offset: The beginning file offset | |
933 | * | |
934 | * Outputs: offset - updated according to number of bytes read | |
935 | * | |
936 | * Returns: The number of bytes sent, errno on failure | |
937 | */ | |
938 | ||
939 | static ssize_t gfs2_sendfile(struct file *in_file, loff_t *offset, size_t count, | |
940 | read_actor_t actor, void *target) | |
941 | { | |
18ec7d5c | 942 | return generic_file_sendfile(in_file, offset, count, actor, target); |
b3b94faa DT |
943 | } |
944 | ||
945 | static int do_flock(struct file *file, int cmd, struct file_lock *fl) | |
946 | { | |
5c676f6d | 947 | struct gfs2_file *fp = file->private_data; |
b3b94faa DT |
948 | struct gfs2_holder *fl_gh = &fp->f_fl_gh; |
949 | struct gfs2_inode *ip = fp->f_inode; | |
950 | struct gfs2_glock *gl; | |
951 | unsigned int state; | |
952 | int flags; | |
953 | int error = 0; | |
954 | ||
955 | state = (fl->fl_type == F_WRLCK) ? LM_ST_EXCLUSIVE : LM_ST_SHARED; | |
956 | flags = ((IS_SETLKW(cmd)) ? 0 : LM_FLAG_TRY) | GL_EXACT | GL_NOCACHE; | |
957 | ||
f55ab26a | 958 | mutex_lock(&fp->f_fl_mutex); |
b3b94faa DT |
959 | |
960 | gl = fl_gh->gh_gl; | |
961 | if (gl) { | |
962 | if (fl_gh->gh_state == state) | |
963 | goto out; | |
964 | gfs2_glock_hold(gl); | |
965 | flock_lock_file_wait(file, | |
966 | &(struct file_lock){.fl_type = F_UNLCK}); | |
967 | gfs2_glock_dq_uninit(fl_gh); | |
968 | } else { | |
969 | error = gfs2_glock_get(ip->i_sbd, | |
970 | ip->i_num.no_addr, &gfs2_flock_glops, | |
971 | CREATE, &gl); | |
972 | if (error) | |
973 | goto out; | |
974 | } | |
975 | ||
976 | gfs2_holder_init(gl, state, flags, fl_gh); | |
977 | gfs2_glock_put(gl); | |
978 | ||
979 | error = gfs2_glock_nq(fl_gh); | |
980 | if (error) { | |
981 | gfs2_holder_uninit(fl_gh); | |
982 | if (error == GLR_TRYFAILED) | |
983 | error = -EAGAIN; | |
984 | } else { | |
985 | error = flock_lock_file_wait(file, fl); | |
986 | gfs2_assert_warn(ip->i_sbd, !error); | |
987 | } | |
988 | ||
989 | out: | |
f55ab26a | 990 | mutex_unlock(&fp->f_fl_mutex); |
b3b94faa DT |
991 | |
992 | return error; | |
993 | } | |
994 | ||
995 | static void do_unflock(struct file *file, struct file_lock *fl) | |
996 | { | |
5c676f6d | 997 | struct gfs2_file *fp = file->private_data; |
b3b94faa DT |
998 | struct gfs2_holder *fl_gh = &fp->f_fl_gh; |
999 | ||
f55ab26a | 1000 | mutex_lock(&fp->f_fl_mutex); |
b3b94faa DT |
1001 | flock_lock_file_wait(file, fl); |
1002 | if (fl_gh->gh_gl) | |
1003 | gfs2_glock_dq_uninit(fl_gh); | |
f55ab26a | 1004 | mutex_unlock(&fp->f_fl_mutex); |
b3b94faa DT |
1005 | } |
1006 | ||
1007 | /** | |
1008 | * gfs2_flock - acquire/release a flock lock on a file | |
1009 | * @file: the file pointer | |
1010 | * @cmd: either modify or retrieve lock state, possibly wait | |
1011 | * @fl: type and range of lock | |
1012 | * | |
1013 | * Returns: errno | |
1014 | */ | |
1015 | ||
1016 | static int gfs2_flock(struct file *file, int cmd, struct file_lock *fl) | |
1017 | { | |
5c676f6d | 1018 | struct gfs2_inode *ip = file->f_mapping->host->u.generic_ip; |
b3b94faa DT |
1019 | struct gfs2_sbd *sdp = ip->i_sbd; |
1020 | ||
b3b94faa DT |
1021 | if (!(fl->fl_flags & FL_FLOCK)) |
1022 | return -ENOLCK; | |
1023 | if ((ip->i_di.di_mode & (S_ISGID | S_IXGRP)) == S_ISGID) | |
1024 | return -ENOLCK; | |
1025 | ||
1026 | if (sdp->sd_args.ar_localflocks) | |
1027 | return flock_lock_file_wait(file, fl); | |
1028 | ||
1029 | if (fl->fl_type == F_UNLCK) { | |
1030 | do_unflock(file, fl); | |
1031 | return 0; | |
1032 | } else | |
1033 | return do_flock(file, cmd, fl); | |
1034 | } | |
1035 | ||
1036 | struct file_operations gfs2_file_fops = { | |
1037 | .llseek = gfs2_llseek, | |
1038 | .read = gfs2_read, | |
18ec7d5c SW |
1039 | .readv = gfs2_file_readv, |
1040 | .aio_read = gfs2_file_aio_read, | |
1041 | .write = generic_file_write, | |
1042 | .writev = generic_file_writev, | |
1043 | .aio_write = generic_file_aio_write, | |
71b86f56 | 1044 | .ioctl = gfs2_ioctl, |
b3b94faa DT |
1045 | .mmap = gfs2_mmap, |
1046 | .open = gfs2_open, | |
1047 | .release = gfs2_close, | |
1048 | .fsync = gfs2_fsync, | |
1049 | .lock = gfs2_lock, | |
1050 | .sendfile = gfs2_sendfile, | |
1051 | .flock = gfs2_flock, | |
1052 | }; | |
1053 | ||
1054 | struct file_operations gfs2_dir_fops = { | |
1055 | .readdir = gfs2_readdir, | |
71b86f56 | 1056 | .ioctl = gfs2_ioctl, |
b3b94faa DT |
1057 | .open = gfs2_open, |
1058 | .release = gfs2_close, | |
1059 | .fsync = gfs2_fsync, | |
1060 | .lock = gfs2_lock, | |
1061 | .flock = gfs2_flock, | |
1062 | }; | |
1063 |